FUNCTIONAL ROLES OF THETA- AND ALPHA-BAND NEURAL OSCILLATIONS IN MEMORY AND ATTENTION

Transcription

1 FUNCTIONAL ROLES OF THETA- AND ALPHA-BAND NEURAL OSCILLATIONS IN MEMORY AND ATTENTION By KRISTOPHER LEE ANDERSON A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA

2 2011 Kristopher Lee Anderson 2

3 To my family: Mom, Dad, and Brother 3

4 ACKNOWLEDGMENTS I would like to start by thanking my mentor, Dr. Mingzhou Ding, whose guidance has given me valuable insight into what it takes to be a successful researcher. His hard work and dedication is a great inspiration and I hope to one day live up to his example and provide similar guidance to future students. Extra special thanks to Dr. Keith Berg for introducing me to cognitive neuroscience research and congratulations on your retirement. I sincerely thank Dr. Kimford Meador for his support and collaboration. Thanks to Dr. Charles Schroeder for sharing the primate data. Thank you to my committee members: Dr. Linda Hermer, Dr. Hans van Oostrom, and Dr. Bruce Wheeler for your support and advice. A very special thanks to Dr. Rajasimhan Rajagovindan for showing me the ropes and being my research big brother through graduate school. I am also very thankful to all of my colleagues in the lab, past and present: Yonghong Chen, Mukesh Dhamala, Yan Zhang, Xue Wang, Anil Bollimunta, Hariharan Nalatore, Sahng Min Han, Mo Jue, Yuelu Liu, Xiaotong Wen, Chao Wang, Haiqing Huang, and Amy Trongnetrpunya for the fun and insightful conversations. I feel comfortable saying that the Ding Lab is the best lab in the BME department. (No offense to the other faculty, though!) The BME staff: Kathryn Whitesides, Tifiny McDonald, Danielle Wise, Anide Pierre- Louis, and April Derfinyak have been awesome. Thank you guys for putting up with my paperwork procrastination. Also, thanks to Art Bautista-Hardman for keeping the computers running smoothly. Thank you to all of the volunteers who participated in our studies and made this research possible, particularly the patients at Shands who gave their time and effort in order to help others during a very difficult time in their lives. 4

5 Above all, I save my most heartfelt appreciation for Carolyn and John Anderson, my mom and dad, for whom I owe everything I am and everything I will be. When times are tough, I draw strength from your unconditional love. When times are good, I look forward to sharing my happiness with you. You guys are literally the best parents in the universe. Also, thanks to my brother, Andy Anderson, for not pantsing me more than you did. 5

6 TABLE OF CONTENTS page ACKNOWLEDGMENTS... 4 LIST OF TABLES... 8 LIST OF FIGURES... 9 ABSTRACT CHAPTER 1 INTRODUCTION Aim Aim Aim ANALYSIS OF CORTICAL THETA RHYTHMS DURING A MEMORY TASK Background and Significance Role of Prefrontal Cortex and Medial Temporal Lobe in Memory Processes Role of the Theta Rhythm in Memory Processes Materials and Methods Subjects and Electrode Placement Experimental Paradigm Data Analysis and Hypothesis Testing Results Behavioral Results Power Results Coherence Results Granger Causality Results Discussion Theta and PFC-MTL Interaction Theta and Neuronal Communication Generation and Propagation of Cortical Theta ATTENTIONAL MODULATION OF THE SOMATOSENSORY MU RHYTHM IN HUMANS Background and Significance Materials and Methods Participants Stimulation Device EEG Recording

7 3.2.4 Experimental Design and Paradigm Source Estimation Data Preprocessing Behavior and Evoked Potential Analysis Spectral Power Analyses Correlation Between Prestimulus Mu Power and Evoked Potential Amplitude Time Frequency Analysis of Mu and Beta Activity in SI Results Behavior Somatosensory Evoked Potential (SEP) Prestimulus Power in 8-12 Hz: Scalp Level Prestimulus Power in 8-12 Hz: Source Level Prestimulus Power in Hz: Source Level From Prestimulus Mu Power To Stimulus Evoked Activity Time-Frequency Analysis of Mu and Beta Activity in SI Discussion Mu and Attention Evoked Activity and Attention Relationship Between Prestimulus Mu and Evoked Activity Summary LAMINAR ANALYSIS OF ELECTROPHYSIOLOGICAL RECORDINGS FROM SI IN NONHUMAN PRIMATES Background and Significance Methods Experimental Task Data Collection Evoked Potential and Current Source Density Analysis Correlation Between Prestimulus Mu Power and P20 Amplitude Phase Realignment and Averaging Results Discussion CONCLUSION LIST OF REFERENCES BIOGRAPHICAL SKETCH

8 LIST OF TABLES Table page 2-1 Total number of bipolar derivations in each area and total number of intergrid pairwise combinations of bipolar signals for each grid pair P-values from Wilcoxon signed-rank test for difference in theta coherence peak values between free recall and baseline

9 LIST OF FIGURES Figure page 2-1 Approximate placement of electrode grids for each of the three subjects Schematic of the experimental paradigm Power spectra in each of the three areas for Subject Inter-grid coherence results for Subject Average percent of significantly coherent inter-grid bipolar signal pairs Mean Granger causality values for PFC-MTL in each subject Representation of the causal relationship for theta between PFC and MTL Schematic of the experimental paradigm Regional sources seeded for source space analysis Somatosensory evoked potential comparison Prestimulus power comparison in the sensor space Power spectral analysis in the source space Mean mu 8-12 Hz band power and beta Hz band power From prestimulus mu power to stimulus evoked response Time course of mu and beta power in SI Stimulus evoked activity in SI From prestimulus mu power to evoked P Hz ongoing activity in SI

10 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy FUNCTIONAL ROLES OF THETA- AND ALPHA-BAND NEURAL OSCILLATIONS IN MEMORY AND ATTENTION Chair: Mingzhou Ding Major: Biomedical Engineering By Kristopher Lee Anderson May 2011 Synchronized neural activity involving widespread networks is common in the central nervous system. This activity often manifests itself as oscillations, which at one point were considered to be background noise or an indication of an idling state of the brain. It is now generally accepted that these oscillations play a role in higher-order cognitive processes, and these roles are currently under active investigation. In this dissertation, we study the roles of theta (4-8 Hz) and alpha (8-12 Hz) band oscillations in two higher-order cognitive processes: memory and attention. First, we studied the role of theta (4-8 Hz) oscillations in the communication between two distant brain regions that are both involved in memory processes. The medial temporal lobe (MTL) and the prefrontal cortex (PFC) are known to be critical structures for human memory processes. Furthermore, it has been suggested that they are part of a memory network. While memory-modulated interaction between PFC and MTL has been observed at the hemodynamic level, it remains unclear what the neuronal process is that mediates the communication between these two areas. Experiments in rodents suggest that field oscillations in the theta band (4-8 Hz) facilitate PFC-MTL interaction. No such evidence has been reported in humans. To address this 10

11 problem, cortical electrical activity from MTL, PFC and lateral temporal lobe was recorded from implanted electrode grids in three epilepsy patients performing a verbal free-recall memory task. The data were analyzed using a parametric spectral method to obtain estimates of power, coherence, and Granger causality. A task-modulated increase in coherence values between PFC and MTL was seen during free recall as opposed to a baseline condition. Concurrently, the number of coherent PFC-MTL site pairs was significantly increased during recall. Granger causality analysis further revealed that the increased coherence is a consequence of higher bidirectional information flow between the two regions, with a generally greater driving from MTL to PFC, namely, (MTL PFC) > (PFC MTL). We then investigated the role of mu and alpha (8-12 Hz) oscillations in somatosensory spatial attention. Neural oscillations with a frequency of around 10 Hz are thought to be a ubiquitous phenomenon in sensory cortices, and it has been hypothesized that the level of 10 Hz activity is related to local cortical excitability. During spatial attention, the visual alpha rhythm has been found to be modulated according to the direction of attention. Specifically, a desynchronization (decrease in amplitude) of the alpha rhythm over visual cortex contralateral to the direction of attention as well as a synchronization (increase in amplitude) over visual cortex ipsilateral to the direction of attention have been reported. These modulations have been associated with both a facilitation and an inhibition of sensory processing, respectively. It has been proposed that the somatosensory mu rhythm serves a similar function to the visual alpha rhythm, and the two rhythms have been found to have similar behaviors in cognitive tasks such as working memory. In this chapter, we investigate whether the somatosensory mu 11

12 rhythm is somatotopically modulated by spatial attention in a way similar to the visual alpha rhythm. 128 channel EEG was recorded while subjects performed a somatosensory spatial attention task. In addition to analyses on scalp recorded data, a spatial filtering method was utilized to investigate spatial attention effects in the source space. The direction of spatial attention was found to have an effect on the ongoing mu rhythm occurring in primary somatosensory cortex as well as stimulus evoked activity. Lastly, an analysis was performed to investigate the correlation between the level of prestimulus mu activity and subsequent evoked activity in primary somatosensory cortex. Finally, we further investigated the previous findings regarding the mu rhythm and its relationship with evoked activity by utilizing microelectrode recordings through the cortical laminae of Area 3b in the primary somatosensory cortex of a rhesus monkey during somatosensory stimulation as well as during a baseline period. We were able to confirm that oscillatory activity in the mu band indeed occurs in primary somatosensory cortex. By examining the stimulus evoked P20 component, a homologue of the human P50 (also known as the P1) somatosensory evoked component, we found evidence supporting the previous interpretations that the human P50 is associated with local inhibition. 12

13 CHAPTER 1 INTRODUCTION Since first being recorded by Hans Berger in 1929 (Berger, 1929), neural oscillations have been a major focus of neuroscience research and there has been much debate as to the functional significance of these oscillations in the brain. Early research led to the theory that prominent neuronal oscillations were merely an indicator of an idling state in the brain. This was due primarily to the fact that the amplitude of these oscillations, particularly the alpha or mu (8-10 Hz) rhythms, appeared to be higher during resting as opposed to active states (Kuhlman 1978; Pfurtscheller et al., 1997). More detailed studies using controlled conditions have led to a rethinking of this view. For example, alpha oscillations recorded over posterior electrodes (Jensen et al., 2002) and theta (4-7 Hz) oscillations recorded over frontal electrodes (Jensen and Tesche, 2002; Meltzer et al., 2008) have both been shown to increase during certain short term memory retention tasks, raising the possibility that these rhythms might play an active part in higher-order cognitive processes. Currently, our understanding of neural oscillations, including both their functional roles as well as the mechanisms of their generation is quickly expanding and it is now becoming clear that they are critical to the proper functioning of the brain, not just epiphenomena of neural firing. This dissertation explores the functional significance of theta- and alpha-band oscillations in two higher-order cognitive processes; namely, memory and attention. This exploration is carried out along three specific aims: Aim 1 An investigation of the theta rhythm as a facilitator of communication between the prefrontal cortex and medial temporal lobe during memory. The medial temporal lobe 13

14 (MTL) and the prefrontal cortex (PFC) are known to be critical structures for human memory processes. Furthermore, it has been suggested that together, they form part of a memory network. While memory-modulated interaction between PFC and MTL has been observed at the hemodynamic level, what remains unclear is the neuronal process that mediates the communication between these two areas. Experiments in rodents suggest that field oscillations in the theta band (4-8 Hz) facilitate PFC-MTL interaction. No such evidence has been reported in humans. We investigated this by recording intracortical EEG while subjects performed a verbal free-recall task and during a relaxed baseline period. Coherence and Granger causality were estimated between prefrontal and medial temporal areas to determine whether or not an interaction exists and, if so, the dynamics of this interaction. The results of this investigation (Chapter 2) have been published (Anderson et al., 2010). Aim 2 An investigation of the effects of directed spatial attention on the somatosensory mu rhythm. Directed attention is a mechanism through which brain resources are allocated to increase detection and/or discrimination of relevant stimuli. Spatial directed attention in the visual domain has been shown to affect the ongoing 8-12 Hz alpha rhythm, specifically causing a suppression of alpha over parieto-occipital sites contralateral to the direction of attention (Sauseng et al., 2005; Thut et al., 2006; Worden et al., 2000; Wyart and Tallon-Baudry, 2008). The somatosensory mu rhythm has traditionally been seen as being functionally similar to the visual alpha rhythm (Pfurtscheller et al., 1996). However, the effect of spatial directed attention on the somatosensory mu rhythm has not yet been reported in the literature. 14

15 In order to investigate the effects of lateralized spatial attention on the somatosensory mu rhythm, we recorded high-density EEG while subjects performed a somatosensory directed attention task. In addition to analyses on scalp recorded data, a spatial filtering method was utilized to investigate spatial attention effects in the source space. Furthermore, we investigate whether mu band oscillatory activity exists in primary somatosensory cortex by performing CSD analysis on ongoing local field potentials recorded directly from Area 3b in an awake, behaving monkey. Aim 3 An investigation of the role of the somatosensory mu rhythm in stimulus processing. Directed spatial attention has been shown to have an effect on early and middle latency SEPs (Eimer and Forster, 2003; Garcia-Larrea et al., 1991, 1995; Jones et al., 2009, 2010; Mauguiere et al., 1997b). Our findings show that directed spatial attention also modulates ongoing somatosensory mu activity. Accordingly, one might suspect that these two phenomena are related in some way. This assumption is supported by experiments showing a relationship between prestimulus mu rhythm amplitude and somatosensory evoked responses. An inverted-u relationship has been found between the amplitude of the ongoing prestimulus mu rhythm and the N1 amplitude (Zhang and Ding 2009). Additionally, mu rhythm amplitude has been shown to influence early evoked responses measured using magnetoencephalography (Nikouline et al., 2000b; Jones et al. 2009, 2010). Using the EEG data recorded while subjects performed a somatosensory attention experiment, we investigated the effect of the amplitude of the somatosensory mu rhythm immediately before stimulus arrival on evoked activity during different attentional states. We follow this up by performing a similar correlation between prestimulus mu 15

16 power and early evoked activity on laminar recordings from an awake, behaving monkey. The results of the human EEG experiement (Chapter 3) have been published (Anderson and Ding, 2011). 16

17 CHAPTER 2 ANALYSIS OF CORTICAL THETA RHYTHMS DURING A MEMORY TASK 2.1 Background and Significance Role of Prefrontal Cortex and Medial Temporal Lobe in Memory Processes Two brain areas consistently implicated in human memory are the medial temporal lobe (MTL) and the prefrontal cortex (PFC). The role of the MTL in memory processes was well established in Scoville and Milner s (1957) work on patient HM, whose bilateral MTL lesions led to anterograde amnesia. Subsequent functional neuroimaging (Cohen et al., 1999; Daselaar et al., 2001) and electrophysiological studies (Fell et al., 2001) have further elucidated the MTL s role in episodic encoding and retrieval. Similarly, lesions in the PFC are known to cause impaired memory functions, such as free recall (Shimamura, 1995; Wheeler et al., 1995) and working memory (Petrides and Milner 1982; Bechara et al., 1998). PET and fmri imaging studies have reported reproducible activations of the PFC in a broad array of memory processes (McIntosh et al., 1997; Fletcher and Henson, 2001; Bunge et al., 2004; Dove et al., 2006). The hypothesis that both structures are part of a unified memory network is supported by the direct anatomical pathways linking PFC and MTL in monkeys and rats (Goldman-Rakic et al., 1984; Squire et al., 1989; Burwell et al., 1995; Suzuki, 1996; Degenetais et al., 2003). In humans, functional connectivity analysis of fmri data has revealed correlations between the lateral PFC and MTL in working memory (Gazzaley et al., 2004), episodic encoding (Grady et al., 2003), and episodic retrieval (Nee and Jonides, 2008). These functional connectivity results suggest that not only are PFC and MTL both involved in memory, their task-related activity is statistically correlated, which can be taken to imply interaction. However, analyses at the hemodynamic level do not provide direct insight 17

18 into how these interactions are mediated physiologically. Additional details of the interaction, including directions of information flow, remain not known Role of the Theta Rhythm in Memory Processes What are the possible neuronal processes that could mediate the interaction between these two areas? Animal studies have long suggested that the theta rhythm, a prominent 4 to 8 Hz oscillatory phenomenon in the limbic system, is closely linked to the formation, storage and retrieval of memory (Miller, 1991; Kahana et al., 2001; Buzsaki, 2002; Vertes, 2005). More recent work postulates that during memory processes, theta oscillations mediate interactions between MTL and other cortical areas (Jensen and Lisman 2005), including PFC (Jones and Wilson 2005). In humans, theta activity has been observed in MTL (Meador et al., 1991; Kahana et al., 1999; Tesche and Karhu, 2000; Eckstrom et al., 2005) as well as in frontal areas (Asada et al., 1999; Jensen and Tesche, 2002; Sederberg et al., 2003). Furthermore, these theta activities were shown to correlate with memory performance. The issue of whether or not theta is a physiological process that mediates PFC-MTL interaction in humans remains unresolved. A recent study by Raghavachari et al., (2006), utilizing human intracranial recordings, found a lack of coherent theta activity between distant cortical sites during working memory. This finding suggests two implications: (1) there is a significant species difference in the processes that mediate cortical interaction during memory or (2) experimental or analytical issues have prevented the observation of the role of theta activity in human inter-areal interaction. To examine this problem, three patients undergoing presurgical evaluation for intractable epilepsy were recruited to perform a free recall memory task. Intracranial electroencephalogram (ieeg), also known as electrocorticogram (ECoG), was recorded 18

19 from multiple implanted electrode grids. The subjects were first given a series of words to remember. Then, following a distraction period, the subjects were asked to recall the words from memory. Data from the period of recall was compared with that from a baseline period of eyes-open fixation. Consistent with the hypothesis that theta acts to mediate memory-related interaction between PFC and MTL, greater theta-band coherence was found during recall between electrodes in the prefrontal and in the medial-temporal areas. Granger causality analysis was used to estimate the directionality of this theta interaction, further defining the role played by each area. 2.2 Materials and Methods Subjects and Electrode Placement Three epilepsy patients gave informed consent and participated in the study. The experimental and recording protocol was approved by the institutional review board of the University of Florida and the affiliated Shands Hospital at the University of Florida. Figure 2-1 illustrates the approximate positions of the implanted electrode grids in each of the three subjects. Two subjects had electrodes placed on the left hemisphere and the remaining subject had electrodes placed on the right hemisphere. All three subjects had grids covering the lateral prefrontal cortex (PFC) and lateral temporal lobe (LTL), as well as strips of electrodes on the ventral surface of the temporal lobe. The two most medial electrodes on the ventral strips, henceforth referred to as subtemporal grids, were treated as proxies of medial temporal lobe (MTL) activity due to their proximity to the parahippocampal region. In each grid, the electrodes were 3mm in diameter with a spacing of 10mm between neighboring electrodes. Depending on the subject, additional grids were implanted, but were not included in the analyses for lack of corresponding coverage in all subjects. 19

20 3.2.2 Experimental Paradigm The experimental paradigm was a verbal free-recall and recognition task. An LCD monitor placed three feet in front of the subject was used to present the stimuli. A fixation cross remained in the center of the screen throughout the experiment. As illustrated in Figure 2-2, the experiment consisted of multiple blocks, with each block starting with the sequential presentation of 20 words (encoding period) chosen from the Kucera and Francis word pool (Kucera and Francis 1967). Each word was displayed on the screen for 2 seconds with a delay between words randomly selected from between 1 and 2 seconds (mean 1.5 seconds). The subject was instructed to try to remember as many of the words as possible. Next, in order to minimize recency effects and to discourage verbal rehearsal, the subjects were asked to count aloud, backward by threes, starting at a random number for 30 seconds (distraction period). Following this distraction period, subjects were given 50 seconds to recall aloud as many of the previously presented words as they could remember (free recall period). After this free recall period, another set of twenty words was presented, which was followed by another distraction period and another recall period. Finally, an additional distraction period was given followed by a presentation of 80 words sequentially on the screen, consisting of the 40 previously presented words along with 40 words which had not been seen previously (recognition period). The subjects were asked to respond by button presses to indicate whether or not they recognized the word as being a previously presented word or a new word. The word stayed on the screen until the subject responded. Each response was followed by a delay of between 1 and 2 seconds (mean 1.5 seconds) before the appearance of the next word. The subjects each performed three blocks. At the end of each block, a 1-minute baseline period 20

21 during which the subject maintained their gaze on the fixation cross was recorded. The subjects were then allowed a short break before beginning the next block. The current analyses were focused on the data from two conditions: free recall and baseline fixation Data Analysis and Hypothesis Testing Data were sampled at 400 Hz by a Nicolet amplifier system, band pass filtered from 0.16 to 30 Hz, and downsampled to 200 Hz. Data segments contaminated by artifacts were excluded from further analysis. The remaining artifact-free data was divided into non-overlapping epochs of 500 ms in length. Epochs from different blocks but within the same condition (free recall or fixation baseline), after a bipolar treatment (see below), were combined and treated as realizations of an underlying stochastic process to be characterized by power, coherence, and Granger causality spectra. Physiological differences in these variables between the two conditions were then assessed. Bipolar derivations: The raw data in each implanted electrode grid were recorded against a common reference electrode fixed to the scalp of the subject contralateral to the hemisphere of grid placement. The reference electrode is not free of neural activity, which may potentially confound functional connectivity measures such as coherence and Granger causality, as the activity underlying this electrode will appear in all recorded channels. Volume conduction presented a further complicating factor. To overcome these problems, the data were re-referenced as bipolar signals. Specifically, for the two most medial electrodes in the subtemporal grid, their difference was treated as a representation of MTL activity. For the frontal and lateral temporal grids, the differences between all pairwise combinations of horizontally, vertically, and diagonally neighboring electrodes were used to represent activity in the respective brain regions. 21

22 These difference signals will henceforth be referred to as bipolar signals or bipolar derivations. Table 2-1 gives the number of such signals in each of the three recording grids for all three subjects. For Inter-grid analysis, three combinations are possible: PFC-MTL, LTL-MTL, and PFC-LTL. Bipolar derivations from every grid were pairwise combined with bipolar derivations from another grid and coherence and Granger causality were computed for each such pair. The total number of such pairs is given in Table 2-1 for each of the three inter-grid combinations in each subject. Spectral analysis: Each pair of bipolar signals was subjected to autoregressive (AR) spectral analysis (Ding et al., 2000; Ding et al., 2006; Rajagovindan and Ding, 2008). Briefly, let the pair of bipolar signals at time t be denoted by X t = (x 1t, x 2t ) T where T stands for matrix transposition. Assume that the data can be described by the following AR model: m A X k 0 E k t k t (Eq ) where E t is a temporally uncorrelated residual error series with covariance matrix Σ, and A k are 2 2 coefficient matrices to be estimated from data (Ding et al., 2000; Ding et al., 2006). The model order m was determined by the Akaike Information Criterion (AIC) (Akaike 1974) and was further verified by comparing the spectral estimates from the AR model with that from the Fourier method. For the data analyzed in this study m = 17 was chosen as a tradeoff between sufficient spectral resolution and over-parameterization. Once the model coefficients A k and Σ are estimated, the spectral matrix can be evaluated according to: 22

23 * S(f) = H( f ) ΣH ( f ) (Eq ) where the asterisk denotes matrix transposition and complex conjugation and m 2 ikf 1 H( f) ( ) k 0 ke A is the transfer function. The power spectrum of channel l is given by S ll (f), (l = 1 or 2), which is the l-th diagonal element of the spectral matrix S(f). The coherence spectrum between channel l and channel k is: C lk ( f) Slk ( f) ( S ( f ) S ( f )) ll kk 1/ 2 (Eq ) with l,k=1,2 and l k. The value of coherence ranges from one to zero, with one indicating maximum interdependence between the two bipolar signals at frequency f and zero indicating no interdependence. The Granger causality spectrum from x 2t to x 1t is defined as (Geweke, 1982; Brovelli et al., 2004; Ding et al., 2006): I ( 22 ) H12( f) 11 ( f) ln(1 ) (Eq ) S ( f) 11 which can be interpreted as the proportion of x 2t s causal contribution to the power of the x 1t series at frequency f. The logarithm is taken to preserve certain favorable statistical properties. Similarly, the causality spectrum from x 1t and x 2t can be obtained by switching the indices 1 and 2 in Eq. ( ). In the present work, Granger causality analysis was carried out for those pairs of bipolar signals whose coherence in the theta range was deemed statistically significant (see below). Interpretation of Granger causality: Statistically, for two simultaneously measured time series, one series can be called causal to the other if we can better predict the 23

24 second series by incorporating past knowledge of the first one (Wiener, 1956). This concept was later adopted and formalized by Granger (1969) in the context of linear regression models of stochastic processes (see Eq. ( )). Specifically, if the variance of the prediction error for the second time series at the present time is reduced by including past measurements from the first time series in the linear regression model, then the first time series can be said to have a causal (directional or driving) influence on the second time series. Reversing the roles of the two time series, one repeats the process to address the question of causal influence in the opposite direction. Here, directions of causal influence are equated with directions of synaptic transmission of neuronal activity (Ding et al., 2006; Bollimunta et al., 2008). Random permutation test for statistical significance: To test the significance of inter-grid coherence, the following procedure was followed (Brovelli et al., 2004), the aim of which is to create a null hypothesis distribution for the peak coherence values in the theta range for both conditions. (1) The 500ms epochs were numbered from 1 to N where N is the total number of epochs for a given condition (free recall or baseline). (2) The epoch index from each brain region was permuted randomly to create a synthetic data set where it is reasonable to assume that no interdependence exists between any pair of bipolar derivations. (3) All pairwise coherence was calculated for the synthetic data set and the single largest peak coherence value among all coherence values in the theta range (4-8 Hz) from all channel pairs was selected. (4) Steps (2) and (3) were repeated many times. (5) The null hypothesis distribution was estimated by fitting an extreme-value distribution to the peak coherence values (Wang et al., 2007). Calculated coherence values from the experimental data were considered significant if they 24

25 exceeded the 99.9th percentile value of the maximum null hypothesis distribution between conditions (p<0.001). The analysis of a frequency range as opposed to that of a single frequency point and the simultaneous comparison of many inter-grid channel pairs present a multiple comparison problem. The traditional Bonferroni correction is not applicable here as the underlying variables may not be independent. Choosing the maximum value in Step (3) is a way to account for this problem (Nichols and Holmes 2002). 2.3 Results Behavioral Results All subjects completed the task according to instructions. Subjects sustained attention throughout the experiment and actively attempted to recall the presented words throughout the free recall periods. The mean percent of words correctly recalled was 27.8±4.7% which is in line with previously reported results in healthy subjects (Fernandez et al., 1998). During the recognition phase of the task, subjects correctly recognized a mean of 92.7±5.5% of the words presented Power Results Spectral power was estimated for all the bipolar signals in each of the three implanted grids using the parametric AR approach and the result from Subject 3 (see Figure 2-1 for approximate electrode locations) is shown in Figure 2-3. A small peak in the range between 4 and 8 Hz was seen in most bipolar derivations, indicating synchronized local theta oscillations (Kahana et al., 1999; Raghavachari et al., 2001; Sederberg et al., 2003; Canolty et al., 2006). The power spectra for the remaining two subjects are similar. The average peak theta frequency between all subjects was: 7.41 Hz for MTL, 7.78 Hz for LTL and 7.35 Hz for PFC. The larger spectral peaks at the 25

26 lower frequencies (~2Hz) are an artifact of combining the band-pass filtering (0.16 to 30 Hz) with the 1/f spectral characteristic of the EEG signal (Slutzky 1937; Pritchard 1992; Demanuele et al., 2007). The thick solid curves are the averages of the bipolar power in each grid for each condition. For all three subjects, no consistent difference in theta power between the two conditions was found in any of the three regions Coherence Results To investigate the level of interaction between the cortical areas, coherence spectra between all possible pairwise combinations of bipolar derivations between a given pair of grids were estimated, and the result for PFC-MTL from Subject 3 is shown in Figure 2-4A. In contrast to the power spectra in Figure 2-3 where the theta peak is rather modest, coherence peaks in the theta range were much more prominent, indicating communication between the two brain regions via theta-band oscillations. Moreover, the average theta coherence (thick solid curves) is higher for free recall than for baseline. This impression is confirmed in Figure 2-4B where peak coherence values in theta range are plotted for the two conditions. The height of the rectangular bars is the mean. A Wilcoxon signed-rank test revealed that the interaction between PFC and MTL was significantly higher when the subject actively recalled words compared to baseline (p= ). No systematic spatial patterns were seen in frontal electrodes that showed increased coherence with MTL. For PFC-LTL the coherence is slightly higher during free recall (p = ). There is no significant difference between the two conditions for LTL-MTL (p = 0.77). This pattern of interaction is found in all three subjects, as summarized in Table 2-2 below. Another way of quantifying the task-related modulation of inter-areal theta synchrony is the number of pairwise bipolar combinations whose peak theta coherence 26

27 values exceed the estimated 99.9% confidence thresholds (horizontal lines in Figure 2-4B; see Methods) which corresponds to a significance level of p< Since the number of channels in each grid varied between subjects, the percentages of pairwise combinations above threshold were calculated and averaged across subjects. As seen in Figure 2-5, there is a highly significant (p = 0.002) task-related increase in the percentage of theta coherent bipolar pairs for PFC-MTL. For PFC-LTL a smaller increase in the number of coherent pairs is observed (p = 0.02). For LTL-MTL, while the percentage of pairs exhibiting significant theta coherence is high, the difference between the two conditions is not significant (p = 0.31) Granger Causality Results Coherence is a symmetric interdependence measure. Namely, when A is coherent with B, B is equally coherent with A. To gain insight into the information flow pattern between the PFC and MTL during memory performance, Granger causality analysis was carried out on the pairs of bipolar signals that showed significant theta coherence during recall. The results for the three subjects are shown in Figure 2-6. The peak and both conditions. This suggests that MTL theta plays a greater role in driving PFC-MTL synchrony than PFC theta. Reciprocal causal influence from PFC to MTL is also seen, indicating that the communication is bidirectional. Consistent with the task-related increase of coherence in Figures 2-4 and 2-5, Granger causality values in both directions show an increase for the free recall condition compared to the baseline fixation condition. For Subjects 1 (left panel) and 3 (right panel) the increase is significant at p<0.01. For Subject 2 (middle panel), the p value cannot be assessed, the 27

28 reason being that no Granger causality value is available during the baseline period due to the fact that the number of bipolar signal pairs exhibiting significant theta coherence for this subject during baseline is zero. These results, schematically summarized in Figure 2-4B, suggest that the increased coherence between the PFC and MTL in Figure 2-2 and Figure 2-3 is a consequence of increased communication in the theta band in both the PFC MTL and MTL PFC directions. 2.4 Discussion In this manuscript, we investigated whether theta oscillations play a role in mediating the interaction between the prefrontal cortex and the medial temporal lobe in human memory processes. Cortical electrical activity from three brain areas (lateral prefrontal cortex (PFC), medial temporal lobe (MTL), and lateral temporal lobe (LTL)) were recorded from implanted electrode grids in three epilepsy patients performing a verbal free-recall task. The multi-electrode data, after a bipolar treatment, were analyzed using an autoregressive spectral method to obtain estimates of spectral power, coherence, and Granger causality. Coherent theta activity was found in all pairwise combinations of the three cortical regions. When the free recall condition was compared to the baseline fixation condition, a large task modulated increase in the overall coherence values between PFC and MTL was seen. At the same time, the number of coherent site pairs between PFC and MTL was also significantly increased during recall. Granger causality analysis of the coherent site pairs further revealed that the increased coherence is a consequence of higher bidirectional information flow between the two brain regions, with a generally greater driving from MTL to PFC. 28

29 2.4.1 Theta and PFC-MTL Interaction The importance of theta activity in the hippocampus and other limbic system structures is well recognized (Buzsaki, 2002; Vertes et al., 2004). Recent experiments with rats have begun to provide evidence in support of the notion that PFC-MTL interaction is mediated through theta oscillations and that this interaction is relevant for memory. Data in rats demonstrate that medial PFC neuronal firings are phase-locked to hippocampal theta and this entrained firing is modulated by animal behavior (Hyman et al., 2005). In addition, increased LFP theta coherence between these two areas has been observed in rats during voluntary behaviors (Young and McNaughton, 2009) and during decision making (Jones and Wilson, 2005). In humans, in spite of extensive evidence implicating PFC and MTL in memory related functions, the questions of whether they work together as part of a network and what physiological processes might mediate their interaction remain unanswered. This lack of understanding could be in part attributable to the difficulty of noninvasive electrophysiological access to MTL structures. Recording of intracranial EEG (ieeg or ECOG) from patients undergoing presurgical monitoring to determine epileptic seizure foci partly overcomes this limitation. A recent study by Raghavachari et al. (2006) used this recording technique to investigate the coherence of theta oscillations between sites throughout the brain during a working memory task. It is reported that significant levels of coherence occurred only between nearby (<20mm) sites, while distant sites very rarely showed coherent theta activity. Consequently, it was concluded that, in different brain areas, cortical theta oscillations are generated independently. In this regard, our study can be seen as the first to report PFC and MTL interaction in the theta band in humans. While the theta coherence values between the two brain areas are generally 29

30 low, and are not statistically significant in most site pairs (> 80%), these values, as well as the number of coherent pairs between PFC and MTL, are nevertheless significantly higher during free recall of remembered words as opposed to a baseline condition. This task-related modulation provides the key evidence for the role of PFC-MTL interaction in memory performance and suggests that theta activity is an underlying physiological process that may mediate this interaction. It should be noted that the memory-related increase in PFC-MTL communication is observed for both hemispheres (see Figure 2-1). While hemispheric asymmetry of memory functions is commonly found (Tulving et al., 1994; Nyberg et al., 1996; Habib et al., 2003), functional imaging studies have shown PFC activation during verbal free recall in both the left and the right hemispheres (Petrides et al., 1995; Fletcher et al., 1998). Hippocampal activation during verbal episodic retrieval has been found bilaterally in multiple imaging studies as well (Lepage et al., 1998) Theta and Neuronal Communication Neuronal ensembles interact and communicate with one another through the transmission of action potentials which carry information. Increased theta coherence during memory recall is a reflection of increased theta phase-locking (Bressler and Kelso, 2001). Siapas et al. (2005) hypothesize that, over short timescales, theta phase locking could be a mechanism for directing information flow between brain regions. They found that neurons constantly fire action potentials at restricted phases of local theta (see also Lakatos et al., 2008). An offset in theta phase between two neurons would allow the neuron with the earlier phase preference to drive the neuron with the later phase preference. Over longer timescales, these consistent relationships would strengthen synaptic connections through spike timing-dependent plasticity. This could 30

31 also lead to the formation of resonant phase-locked loops between regions with activation and transmission delays summing to around 150ms, which corresponds to the estimated delays between PFC and MTL in humans (Miller, 1991). Along another line of reasoning, Jensen (2001) suggests that cortical theta activity that is synchronous, yet out of phase, with hippocampal theta could allow decoding of phase-encoded hippocampal output. This phase decoder could be driven by neurons which fire out of phase, yet entrained to the hippocampal theta rhythm. With respect to the present experiment, two other considerations are relevant. First, field potential oscillations are accompanied by rhythmic bursts of action potentials. Lisman pointed out that bursting is a more reliable means of transmitting information over long distance than single action potentials (Lisman, 1997). Second, Lengyel et al. (2005), using computational modeling, show that memory retrieval might occur in a theta rhythmic fashion. Thus, the idea of theta serving as information carrier during memory recall can be seen as grounded in both theoretical and empirical considerations. This is in further agreement with proposals where the role of theta oscillations in facilitating communication between the cortical and MTL structures has been emphasized (Miller, 1991; Jensen, 2005; Johnson, 2006) Generation and Propagation of Cortical Theta Extensive evidence in animal studies as well as the evidence presented here for humans suggests that theta oscillations may facilitate communications between PFC and MTL. The question remains as to where these oscillations are generated and how they propagate. The hippocampal theta rhythm is generally believed to be caused by input from rhythmically bursting GABAergic and cholinergic neurons in the medial septum (Vertes and Kocsis, 1997) as well as through recurrent connections within the 31

32 hippocampus proper (Kocsis et al., 1999). The mechanisms underlying the generation of cortical theta are less understood. Cortical neurons have the ability to generate theta activity through the action of GABAergic interneurons (Blatow et al., 2003). Cholinergic input from structures in the basal forebrain is another important contributor (Liljenstrom and Hasselmo, 1995; Jones, 2004). Using phase analysis in rats, Siapas et al. (2005) proposed that PFC theta is the result of unidirectional theta input from the hippocampus. This idea is supported by Tierney et al. (2004), who found that hippocampal activity directly influences prefrontal interneurons which, as mentioned above, can create rhythmic cortical theta activity. These results seem to imply a passive role for the PFC in memory performance, a viewpoint at variance with work postulating an active role of the PFC in both memory encoding (Fletcher et al., 1998a) and retrieval (Buckner and Wheeler, 2001, Fletcher et al., 1998b). Our Granger causality analysis shows that the PFC-MTL interaction is bidirectional and, during free recall of verbal information, the compared to a baseline condition, with a generally greater causal influence from MTL to TL). This result supports an active role for the PFC in the present experimental paradigm. It is also in agreement with the prevailing notion that, at the top of the executive control hierarchy, the PFC coordinates posterior brain areas for goal oriented behavior (Knight et al., 1999; Fuster, 2001; Miller and Cohen, 2001). It should be cautioned, however, that the bidirectional PFC-MTL interaction inferred from Granger causality, while consistent with the recurrent anatomical pathways existing between the two areas, cannot rule out the possibility that a third structure drives both 32

33 PFC and MTL (Kaminski et al., 2001). Further investigations with more extensive spatial sampling are needed to elucidate the exact network mechanism. 33

34 Table 2-1. Total number of bipolar derivations in each area and total number of intergrid pairwise combinations of bipolar signals for each grid pair. Subject # Number of bipolar derivations Number of inter-grid combinations PFC LTL MTL PFC-MTL PFC-LTL LTL-MTL Table 2-2. P-values from Wilcoxon signed-rank test for difference in theta coherence peak values between free recall and baseline. Subject # Grid pairs PFC-MTL PFC-LTL LTL-MTL

35 Figure 2-1. Approximate placement of electrode grids for each of the three subjects. 35

36 Figure 2-2. Schematic of the experimental paradigm. 36

37 Figure 2-3. Power spectra in each of the three areas for Subject 3. Gray curves: spectra from individual bipolar derivations. Thick curves: average spectra. The large peak seen at the low frequency range (~2Hz) is an artifact resulting from combining the high-pass action of the band-pass filter (0.16 to 30 Hz) with the 1/f type spectral characteristic of EEG data. Theta frequency range is indicated by shaded background. 37

38 Figure 2-4. Inter-grid coherence results for Subject 3. A) Coherence spectra of all pairwise combinations of bipolar derivations between PFC and the posterior MTL electrodes (gray curves) (see Figure 1). Thick curves: averages of the gray curves. Theta frequency range is indicated by shaded background. B) Coherence peak values in the theta range. Individual peaks are plotted as black dots. The horizontal placements of the dots within each bar are random. From left to right: PFC-MTL, PFC-LTL, and LTL-MTL. Each bar represents the mean of the peak coherence values in one condition with B denoting baseline and R denoting recall. The gray horizontal lines indicate the significance threshold corresponding to p=0.001 for different grid pairs. 38

39 Figure 2-5. Average percent of inter-grid bipolar signal pairs whose theta coherence exceeds the significance threshold for both baseline (white bars) and free recall (black bars) conditions. The standard errors are plotted as error bars. The one-sided t-test p values are included below each plot. 39

40 Figure 2-6. Mean Granger causality values for the coherent pairs of bipolar signals for PFC-MTL in each subject. The number of such pairs is given below each condition. White and black bars represent MTL PFC and PFC MTL respectively. Standard errors are plotted as error bars. Conditions where MTL PFC is larger than PFC MTL at p=0.05 level using a one-sided paired t-test are marked by an asterisk. Such comparison was not done for the baseline condition for Subject 2 as no coherent bipolar pairs were found for that subject in the baseline condition. Results from both subtemporal grids are combined for Subject 3. 40

41 Figure 2-7. Schematic representation of the causal relationship for theta activity between PFC and MTL. 41

42 CHAPTER 3 ATTENTIONAL MODULATION OF THE SOMATOSENSORY MU RHYTHM IN HUMANS 3.1 Background and Significance Field oscillations in the 8-12 Hz range have been observed in visual, somatosensory, and auditory cortices, and they are referred to as alpha, mu, and tau rhythms, respectively. Among these rhythms, visual alpha is the most extensively studied (Shaw, 2003). Typically, the amplitude of alpha is higher over visual areas that are not engaged in a task and lower over areas that are engaged. In particular, when attention is deployed to a location in visual space, decreased alpha amplitudes (desynchronization) have been found over visual areas contralateral to the direction of attention (Sauseng et al., 2005; Thut et al., 2006; Wyart and Tallon-Baudry, 2008), a modulation which is thought to reflect the engagement of relevant cortical areas (Medendorp et al., 2007) through a local increase of cortical excitability (Klimesch et al., 2007; Romei et al., 2008; Worden et al., 2000). Conversely, an increase in alpha amplitude, known as synchronization, has been reported over visual cortex ipsilateral to the attended direction (Kelly et al., 2006; Rihs et al., 2007; Worden et al., 2000; Yamagishi et al., 2003). This increase is considered by some to reflect a gating mechanism, whereby processing of irrelevant stimuli is inhibited in order to better process relevant stimuli (Cooper et al., 2003; Jensen et al., 2002; Klimesch et al., 2007). It has been shown that alpha synchronization and desynchronization also play an important role in other high-order cognitive processes such as memory and visual imagery (Jensen et al., 2002; Klimesch et al., 1999; Medendorp et al., 2007; Tuladhar et al., 2007). 42

43 Are these functional properties of visual alpha shared by similar oscillations in other sensory cortices? The mu rhythm over somatosensory cortex (Gastaut, 1952) is known to behave similarly to the visual alpha rhythm in some respects. For example, a decrease in amplitude of mu, known as an event related desynchronization (ERD), has been noted following somatosensory stimulation (Pfurtscheller, 1989; Nikouline et al., 2000a; Della Penna et al., 2004), which is akin to the parieto-occipital alpha ERD that occurs after visual stimulation (Pfurtscheller et al., 1979; Vijn et al., 1991; Pfurtscheller et al., 1994). As another example, alpha rhythms overlying dorsal stream ( where pathway) areas have been found to increase in amplitude during a visual working memory task that engages the ventral stream ( what pathway) (Jokisch and Jensen, 2007). A similar effect has been seen during a somatosensory delayed-match-tosample task, where mu power was higher over areas hypothesized to be not engaged by the task, such as somatosensory cortex ipsilateral to the sample stimulus (Haegens et al., 2010). Despite these similarities, whether and how the ongoing somatosensory mu rhythm responds topographically to spatial attention, a hallmark of visual alpha reactivity, has only recently been investigated. Jones et al. (2010), using magnetoencephalography (MEG), found that cued spatial attention to the hand decreased mu power in the hand area of SI while attention to the foot on the same side of the body was accompanied by a mu power increase in the hand area. Also utilizing MEG, van Ede at al. (2010) found that attentive as well as non-attentive expectation of a somatosensory stimulus modulated, in a lateralized manner, the beta rhythm, a Hz somatomotor rhythm with some functional similarities to the mu rhythm (Salenius et 43

44 al., 1997; Pfurtscheller and Lopes da Silva, 1999; Ritter et al., 2009). No such expectation effect was seen for the mu rhythm. While in the MEG modality, beta-band activity is often analyzed along with or in lieu of mu activity (Jones et al., 2009; Jones et al., 2010; van Ede et al., 2010), the beta rhythm is often not very prominent in EEG recordings (Zhang and Ding, 2010). The effect of lateralized spatial attention on the somatosensory mu rhythm currently remains uninvestigated. We recorded high-density EEG while subjects performed a somatosensory spatial attention task in which sustained attention was directed to either the right or the left hand. Oscillatory activity in the 8-12 Hz (mu) and Hz (beta) frequency range during a prestimulus time period when somatosensory attention was deployed to either direction was measured and compared with a baseline period. A spatial filter was applied to the scalp-recorded data in order to investigate the effects of somatosensory spatial attention on cortical areas such as primary somatosensory cortex (SI), secondary somatosensory cortex (SII), posterior parietal cortex, lateral and medial frontal areas, and occipital cortex. The inclusion of the occipital cortex (a) removed a potential source of volume conduction that might negatively impact the estimation of somatosensory mu activity which is smaller in magnitude than visual alpha and (b) allowed the examination of possible cross-modal attention effects. Finally, the effect of the mu rhythm prior to stimulus onset on stimulus processing during different states of attention was investigated by correlating prestimulus mu power with evoked activity in SI. Although the relationship between prestimulus mu power and evoked potentials has been investigated previously (Nikouline et al., 2000b; Jones et al., 2009, 2010; Reinacher et al., 2009; Zhang and 44

45 Ding, 2010), the current analyses extend these findings by revealing the impact of source-localized estimates of high, low, and intermediate amplitudes of prestimulus mu on early and late evoked activity during different attentional states. 3.2 Materials and Methods Participants A total of 15 healthy right-handed subjects (Aged years, 8 female) participated in the experiment. All participants provided written informed consent and were paid in accordance with the guidelines of the Institutional Review Board (IRB-02) at the University of Florida. All subjects performed the task according to instructions and were included in the following analyses Stimulation Device Somatosensory stimuli were delivered using a two-channel, custom built, computer controlled, constant-current stimulation device. The device was optically isolated from the stimulus presentation computer and battery powered to ensure the participant s safety and that no additional line noise was introduced into the recording. Stimulus amplitude was adjustable from 0 to 5 ma in ~0.02 ma steps and stimulus duration was fixed at 0.5 ms. Event triggers sent to the EEG recording amplifier were precise to the sub-millisecond level EEG Recording The experiment took place in a dimly-lit, acoustically and electrically shielded booth. Subjects sat comfortably in a chair with their arms apart and resting on a table in front of them. They were instructed to keep their eyes open and fixated on a small cross on a computer monitor 1.5 m in front of them throughout the experiment. 45

46 The EEG data were acquired using a 128-channel BioSemi ActiveTwo System ( with a sampling rate of 2048 Hz. Four channels of electrooculogram (EOG) were recorded in addition to the 128 scalp channels. Statistical analyses of scalp-recorded data were performed on electrodes CP3 and CP4. These electrodes were chosen to represent activity over primary somatosensory cortex because they are where the largest early (~50ms post-stimulus) evoked activity was measured Experimental Design and Paradigm The task was a somatosensory oddball task involving directed spatial attention. A block design was used. Subjects were instructed to fixate on a cross in the center of a computer screen and direct their attention to either their right (attend-right or ATTR), left (attend-left or ATTL), or both (attend-both or ATTB) hands during a block (Figure 3-1). Each block consisted of 70 electrical stimuli being delivered over either the right or left median nerve with equal probability. The inter-stimulus interval was uniformly distributed between 2.5 and 3.5 seconds. Each stimulus could either be a standard (low amplitude, 80-92% probability) or a target (higher amplitude, 8-20% probability). The amplitude of the standard stimulus was held fixed throughout the experiment at twice the detection threshold for each hand. Here, the detection threshold for each hand was determined using an up-down staircase procedure (Leek, 2001) to find the amplitude at which the subject detected the stimulus 50% of the time. Amplitudes for the target stimuli were initially set during a short practice run before the experiment to achieve a target detection error rate of around 25% for both the attend-left and attend-right conditions. During the experiment, target amplitudes were adjusted at every third block to keep the error rate for the attend-left and attend-right conditions around 25%. As the target 46

47 stimuli were not held constant throughout the experiment, only data from standard stimuli were used for the analyses in this paper. Before each block, the subjects were instructed to mentally count the number of target stimuli delivered to the attended hand and to ignore stimuli delivered to the unattended hand. In the attend-both condition, the subjects were instructed to count the total number of targets delivered to both hands. The subjects verbally reported the number of detected targets at the end of each block. A fourth, baseline, condition without any stimuli was recorded at the beginning of the experiment and every six blocks, in which the subject was instructed to relax and stare at the fixation cross (as in the other blocks) for 3 minutes. The order of the blocks, in groups of three, alternated between ATTB ATTR ATTL and ATTB ATTL ATTR. In total, the experiment consisted of between blocks of stimuli (5-6 each of ATTB, ATTR, ATTL) and 3-4 baseline blocks, resulting in stimuli per-condition, per-hand ( total). For the current study, only the attend-left, attend-right, and baseline conditions were analyzed Source Estimation Electrode locations, as well as three fiducial landmarks, were digitized by means of a Polhemus spatial digitizer. Regional dipole source analysis (Scherg, 1992) was used to create a spatial filter using the Brain Electrical Source Analysis (BESA) software package which implements a least-squares algorithm to solve the overdetermined problem and estimate the activity contributed by each source to the scalp-recorded data. Based on findings from previous research, relevant fixed regional sources were seeded into a 4-shell ellipsoidal head model (brain, CSF, skull, and skin conductivities 47

48 of.33, 1.0,.0042, and.33 mohm/m, respectively) and source activity was estimated from each subject s continuous scalp data for further analyses. As illustrated in Figure 3-2, 11 sources were seeded in relevant brain areas: Bilateral primary somatosensory (SI) sources were seeded near the postcentral gyrus, consistent with the hand area found in previous studies (Valeriani et al., 1997; Waberski et al., 2002; Bowsher et al., 2004; Della Penna et al., 2004; Gaetz and Cheyne, 2006; Ritter et al., 2009). Bilateral secondary somatosensory (SII) sources were seeded near the parietal operculum. Coordinates were chosen based on a meta- analysis (Eickhoff et al., 2006). Bilateral posterior parietal (PP) sources were seeded near the superior parietal lobule, a location indicated as being involved in maintaining spatial attention (Corbetta et al., 1998; Kastner and Ungerleider, 2000). Bilateral lateral-frontal (LF) sources were seeded near the middle frontal gyrus. Source sensitivity maps (not shown) for these dipoles indicated contributions from both dorsolateral prefrontal cortex and frontal eye fields. A medial frontal source (MF) was seeded near the inter-hemispheric space between the left and right anterior cingulate gyri. Bilateral occipital (O) sources were seeded near the foveal confluence, an area where V1,V2, and V3 are thought to converge (Dougherty et al., 2003; Schira et al., 2009). In addition to the above 11 task-relevant sources, 5 more sources were seeded to minimize contamination of estimated source activity from other brain areas. Bilateral sources were seeded near the frontal poles. These sources were used to account for ocular activity that was below the rejection threshold. Central and parietal midline sources were seeded to minimize lateral source sensitivity overlap. A deep midline source was seeded to account for additional brain activity. The source sensitivity map (not shown) indicated mostly local (subcortical) and inferior temporal lobe contributions to this source. For each of the 16 regional sources seeded, magnitudes of ERPs and spectral estimates of ongoing neuronal activities from the three dipolar components were used 48

49 to obtain orientation-independent measures. Note that, as no structural MRI was recorded from the participants, source locations should be considered approximate. However, estimated source waveforms are relatively insensitive to variations in dipole location Data Preprocessing Two sets of data, sensor-level and source-level, were analyzed in this study. Preprocessing steps were similar for both data sets, and any differences will be noted below. As the exact locations of the recording electrodes were slightly different for each subject, spherical spline interpolation was used to transform each subject s channels of data into a standard 81-channel montage. This spatially interpolated data set was used for all sensor-level analyses. Channels with poor signal quality for each individual subject were not included in this transformation. All sensor-level analyses were performed on the average referenced data. First, the signals were band-pass filtered between 0.3 and 85 Hz and downsampled to 256 Hz for subsequent analyses. The data were then epoched around each standard stimulus from -700 ms to 500 ms. For baseline data, artificial triggers were inserted into the continuous recordings every 600 to 800 ms, and epoched as above (note that only the prestimulus period from -500 ms to 0 ms, with 0 ms denoting the onset of an artificial trigger, was used for baseline analyses). After this, the DC component was subtracted from each epoch. Any epoch with activity in the EOG channels exceeding 75 uv, or with activity exceeding 50 uv in any scalp channel, was excluded from further analysis. This procedure resulted in between ~15% to ~30% of epochs being rejected from each subject. 49

50 3.2.7 Behavior and Evoked Potential Analysis Behavioral performance for each block was measured as (targets reported-actual targets)/(actual targets). The amplitude of target stimuli was adjusted throughout the experiment to obtain consistent behavioral results. The mean of the prestimulus baseline period from -100 to 0 ms was subtracted from each epoch before averaging. The ERPs for each subject were weighted equally to compute the grand average. The source space ERPs were computed as the square root of the sum of each of the three dipole components squared. A Wilcoxon signedrank test was used at each time point to test whether the difference between conditions was statistically significant. If the tests on at least three consecutive sample points (~12 ms) resulted in p-values less than 0.05, the effect in that time period was considered significant Spectral Power Analyses Spectral power analyses were performed on a time period immediately preceding all artifact-free standard stimuli in the attend-right and attend-left conditions. For the baseline condition, the analyses were performed on the same time window preceding the artificially inserted triggers described in Section This time window was defined to be from -500 ms to 0 ms relative to each stimulus/trigger. This prestimulus window was chosen to be short enough as to minimize the effect of the neuronal response to the previous stimulus and to capture a stable state of the brain at the time of stimulus onset while at the same time long enough to allow for sufficient frequency resolution. For all channels and source dipoles, power spectral densities (PSDs) were estimated for each prestimulus epoch using a multitaper FFT approach with 3 DPSS tapers over this time window, resulting in ±4 Hz smoothing. In order to obtain an orientation- 50

51 independent measure in the source space, magnitudes of the PSDs for each source at each epoch were obtained by taking the square root of the sum of the squares of each dipole component s PSD. When plotting power spectra (Figures 4 and 5), spline interpolation was used to obtain a more smooth curve. We used the estimated spectral power between 8 and 12 Hz to compute the average mu band power and 15 to 35 Hz to compute the average beta band power. To facilitate averaging across subjects, the average band power was normalized by finding the ratio between the power of each condition and the mean power of all three conditions for that subject. As an example, the normalized band power for the ignore condition in a single subject would be calculated as: (ignore)/(ignore+attend+baseline)/3. A Wilcoxon signed-rank test was used to test whether the difference between two conditions was statistically significant Correlation Between Prestimulus Mu Power and Evoked Potential Amplitude To analyze the correlation between prestimulus mu power and evoked potential amplitude in SI, trials for each subject and each condition were divided into two groups: right stimuli and left stimuli. The trials in each group were then rank ordered by the amplitude of the prestimulus mu power estimated from the SI source contralateral to stimulation, and sorted into 5 bins of equal size with an overlap of 50%. Each bin contained about 33% of the total available trials in each group. The power bins were indexed from 1 to 5 where Bin 1 has the smallest mu power and Bin 5 has the largest. For each subject, the trials within a power bin were used to calculate the evoked activity in source SI in the same way as described in the above section on behavior and evoked potential analysis. The mean amplitudes of the evoked activity in two time ranges: 45 to 55 ms and 140 to 160 ms, centered on the peaks of evoked activity in the 51

52 SI source, and where significant differences between the amplitudes of the sensor-level SEP were found, were then computed for each bin. To minimize the effect of intersubject variability in evoked activity amplitude on population averaging, the following procedure was adopted to normalize the data from each subject. Let the mean amplitude for Subject K in Power Bin J be denoted as A(K,J). The mean evoked amplitude for this subject will be calculated as mean_a(k) = [A(K,1) + A(K,2) + + A(K,5)]/5. The normalized evoked amplitude was calculated as the percent change against this mean, namely, norm_a(k,j) = (A(K,J) mean_a(k))/mean_a(k). This normalized evoked amplitude was then averaged across subjects to obtain the mean normalized evoked amplitude for each power bin. The results were then combined across hemispheres. For the early evoked component (45 to 55 ms), Spearman s rank correlation coefficients were computed to test for statistical dependence. A quadratic regression was performed on the late component (140 to 160 ms). Similar analyses were performed on the other bilateral sources (SII, PP, LF, and O). For each source, the time intervals chosen for analysis were defined according to where a significant difference was found between the grand average evoked activity to attended and unattended contralateral somatosensory stimuli Time Frequency Analysis of Mu and Beta Activity in SI The temporal evolution of mu and beta power was compared between different attention conditions for the SI sources. First, the data were epoched from to 2500 ms around each standard stimulus or, for the baseline condition, each artificially inserted trigger (see Section 3.2.6). Epochs containing artifacts during this time period were rejected from further analysis. The data from each epoch were then timefrequency decomposed by convolving with Morlet wavelets to obtain power estimates 52

53 from 8-12 Hz and Hz with center frequencies at 1 Hz intervals. The time course of the mean band power was then calculated for each trial and averaged within each condition. As with the prestimulus power analysis described in Section 3.2.8, an orientation-independent measure was obtained by calculating the magnitude of the power estimates from the three dipoles in each regional source for each time point. Results were then combined across hemispheres in a way similar to that described in Section 3.2.8, with the additional step of using the temporal mean of each condition (as opposed to the power at each time point) as the normalization value. As an example, the normalized mu power time course for the ignore condition in a single subject would be calculated as: Normalized_ignore(t) = ignore(t)/(temporal_mean(ignore)+temporal_mean(attend)+temporal_mean(baseline))/3. A Wilcoxon signed-rank test was used to test whether the difference between two conditions was statistically significant at each time point. 3.3 Results Behavior All 15 subjects performed the task according to instructions. The error-rate in target detection averaged across subjects for each condition was: 25.6% (±2.3%) for attend-left and 25.5% (±1.9%) for attend-right. This rate was maintained throughout the experiment by adjusting the amplitude of the target stimuli to ensure consistent task difficulty in both the attend-left and attend-right conditions. The amplitudes of standard stimuli, which were held fixed at twice the detection threshold for each subject throughout the experiment, averaged across subjects were 2.71 ma (±0.17 ma) for right standard stimuli and 2.58 ma (±0.14 ma) for left standard stimuli. The generally higher detection threshold for the right hand, consistent with 53

54 previous reports (Friedli et al., 1987; Meador et al., 1998), reflects handedness-related threshold asymmetry (all subjects in the current study were right-handed) Somatosensory Evoked Potential (SEP) The grand average SEP waveforms for stimuli delivered contralaterally and ipsilaterally to recording electrodes over somatosensory cortex (CP3 and CP4) under attend and ignore conditions are shown in Figure 3-2A and B. Data from the two hemispheres have been combined. The P1 component, sometimes also referred to as the P45, P50, or P60 component, peaks around 50 ms and is only seen in the hemisphere contralateral to stimulation. This component is significantly larger for the ignore condition compared with the attend condition. The N1 component, a bilateral negative component peaking around 150 ms and sometimes referred to as the N140, shows the opposite effect; a greater amplitude for attended stimuli than ignored stimuli. For contralateral stimuli, this negativity extends from 150 ms to 200 ms, overlapping a central positive component that peaks around 200 ms. Figure 3-2C shows the grand average SEP for all stimuli (right and left) under attend and ignore conditions which emphasizes the bilateral N1 attention effect. The N1 effect is slightly larger in the right hemisphere as opposed to the left, as seen in Figure 3-2D, which is a topographic plot of the mean difference between all attended stimuli (right stimulus, attend right; left stimulus, attend left) and all ignored stimuli (right stimulus, attend left; left stimulus, attend right) in the time period from 140 to 160 ms. This effect is consistent with a right hemispheric dominance in the parietal lobes during spatial attention (Heilman and Abell, 1980; Heilman et al., 1985; Mesulam, 1999; Meador et al., 2002). 54

55 3.3.3 Prestimulus Power in 8-12 Hz: Scalp Level In the period prior to stimulus onset, oscillatory activity can be used as a measure to give insight into the state of the brain and how directed attention modulates this state to facilitate information processing. Figure 3-3 shows the effects of somatosensory attention on oscillations in the mu band (8-12 Hz) recorded at the scalp level. Normalized prestimulus power spectra for sensors CP3 and CP4, averaged over all subjects, are plotted in Figure 3-3A. A peak in the mu band (8 to 12 Hz) exists in both sensors for all conditions. It can also be seen in both sensors that the average prestimulus mu power measured over somatosensory cortex contralateral to the direction of attention is lower than the power over cortex ipsilateral to the direction of attention. A one-sided Wilcoxon signed-rank test of the difference in average mu power between attend-ipsilateral and attend-contralateral resulted in p=0.11 and p=0.06 for CP3 and CP4, respectively. Figure 3-3B shows the average percent difference in 8-12 Hz band power between the attend-right and attend-left conditions over the entire scalp. Consistent with Figure 3-3A, it can be seen that when somatosensory attention is directed to the right side, mu power over the contralateral (left) somatosensory cortex is lower than the mu power in the ipsilateral (right) somatosensory cortex. The effect appears to be localized to sensors lying over somatosensory cortex. Similar patterns of alpha power reduction have been observed over visual cortex with visual spatial attention (Thut et al., 2006; Rajagovindan and Ding, 2010). It is worth noting that while an attention-related decrease in mu power was seen in Figure 3-3A, the difference was not highly significant (p=0.11 for CP3 electrode and p=0.06 for CP4 electrode). It is likely that, given the large visual alpha activity, the mu 55

56 power estimation is adversely affected by volume conduction from the occipital cortex, which may also explain the broad increase in 8-12 Hz power during somatosensory attention compared with baseline in brain areas outside the somatosensory cortex in Figure 3-3C. This problem is overcome below by carrying out spectral power analysis in the source space Prestimulus Power in 8-12 Hz: Source Level All power results in the following sections are obtained from the magnitude of the PSDs of the three components of each regional source dipole (see Methods). Figure 3-5 shows the mean power spectra for the regional sources collapsed across conditions and hemispheres. Spectral power peaks in the 8-12 Hz range can be seen in the somatosensory (SI, SII), posterior parietal (PP), and occipital (O) sources for all 15 subjects. A slight peak in the beta band (~20 Hz) can be seen in the SI and SII sources for 2 subjects and 3 subjects, respectively. The frontal sources, lateral frontal (LF) and medial frontal (MF), do not show peaks in the mu frequency range, though a slight bump in the theta range can be seen in some subjects. Spectral power estimates from 0 to 3 Hz are not plotted, as combining the high-pass action of the bandpass filtering with the 1/f spectral characteristic of the electroencephalography signal can create an artificial spectral peak in this frequency range (Slutzky, 1937; Demanuele et al., 2007). The largest oscillations in the 8-12 Hz frequency band occur in the occipital sources, where the grand average peak is 6.5 V2/Hz, compared with peaks of 2.9 V2/Hz, 2.3 V2/Hz, and 2.1 V2/Hz in SI, SII, and PP sources, respectively. If the somatosensory 8-12 Hz oscillations were due to voltage propagation from the occipital cortex, one would expect the amplitude of the oscillations measured from the posterior parietal sources (located between somatosensory and visual cortices) to lie between 56

57 the amplitudes of the somatosensory and occipital oscillations. This is not the case, as the peak amplitude in the posterior parietal sources is less than the somatosensory sources as well as the occipital sources. Therefore somatosensory mu oscillations appear to be generated in local cortices. The 9.0 Hz peak seen in SII could represent the sigma rhythm, a 7-9 Hz rhythm recorded in SII that is responsive to somatosensory stimulation (Narici et al., 2001), though voltage propagation from SI cannot be ruled out. One phenomenon to note is that mu activity was found in the SI sources of all 15 subjects. This finding contrasts with earlier reports where the incidence of observable mu oscillations in scalp EEG varied from 4% to 60% (Shaw, 2003). However, Niedermeyer (1997) hypothesized that the mu rhythm is a universal phenomenon in healthy adolescents and adults and McFarland et al., (2000) report that mu activity can be detected in most normal adults through spectral analysis of EEG activity. Figure 3-6 (top) shows the results of a comparison of mu band power between conditions in the source space. For a given source, the condition where attention is directed contralaterally to the source hemisphere is designated as the attend condition, and the condition where attention is directed ipsilaterally to the source hemisphere is designated as the ignore condition. The results are combined across hemispheres and band-power values for each subject were normalized according to the procedure described in Section For the primary somatosensory (SI) sources, a significant decrease in mu power is seen in the attend condition when compared with the ignore condition (p=0.002) and baseline (p=0.031). No significant difference is seen between the ignore and baseline 57

58 conditions in SI (p=0.35). In addition, 8-12 Hz (possibly sigma) oscillations in SII do not appear to be modulated by the current spatial attention task (p>0.1 between all conditions), although past work has shown that attention modulates stimulus-evoked responses in SII (Hsiao et al., 1993; Mima et al., 1998; Steinmetz et al., 2000; Fujiwara et al., 2002; Hoechstetter et al., 2002). No significant differences are found between conditions in the posterior parietal sources. Power in the 8-12 Hz band is not significantly modulated in lateral frontal cortex by the current spatial attention task (p>0.1 between all conditions). The medial frontal source was not included in this analysis as it has no laterality, and thus attend contralateral and attend ipsilateral are undefined in the prestimulus period for this source. In occipital cortex, a significant power increase is seen from baseline to somatosensory attention conditions (p= for baseline vs ignore and p= for baseline vs attend). This is consistent with the gating hypothesis that an increase in 10 Hz oscillations occurs over cortical areas that are irrelevant to the task (Foxe et al., 1998; Jensen et al., 2002; Cooper et al., 2003; Rihs et al., 2007; Klimesch et al., 2007) Prestimulus Power in Hz: Source Level According to a previous MEG study (van Ede et al., 2010) beta band is defined to be Hz. As shown in Figure 3-6 (bottom) prestimulus beta power during the attend condition is significantly less than during the ignore condition (p = 0.048) in the primary somatosensory (SI) cortex. Attention does not appear to significantly modulate beta oscillations in any of the other cortical areas. The larger error bars (standard errors of the mean) are likely due to the lack of precision in estimating beta power in the absence 58

59 of consistent spectral peaks in this frequency range. Further analyses below will focus on the mu band, where the attention effects are more robust From Prestimulus Mu Power To Stimulus Evoked Activity The above results showed that, over primary somatosensory cortex contralateral to the attended direction, attention reduced mu power prior to stimulus onset and at the same time modulated the stimulus-evoked response. Presumably, the prestimulus mu power reduction contributed to the subsequently improved stimulus processing. In order to investigate the relationship between pre- and post-stimulus activity, we estimated the magnitude of the evoked potential in SI for two time periods: 45 ms to 55 ms (early), and 140 ms to 160 ms (late) as a function of different levels of prestimulus mu power in the same hemisphere under the attend condition (attend contralaterally to source hemisphere) and the ignore condition (attend ipsilaterally to source hemisphere) (see Section for details). Stimuli delivered contralateral to each SI source were included and results were combined across hemispheres. As seen in Figure 3-7, a significant positive linear relationship was found between prestimulus mu power and the early evoked component for both attended and ignored stimuli (Spearman rank correlation rho=0.26, p=0.037 for attended stimuli and rho=0.29, p=0.019 for ignored stimuli). The later component followed a nonlinear quadratic relationship with prestimulus mu power in both conditions. For attended stimuli, the relationship was of an inverted U type, with a p value for the F statistic of The relationship for ignored stimuli followed an upright U shape with p= The spatial specificity of the relationship between prestimulus mu power and evoked response was investigated by employing a similar analysis for the remaining sources. For each source, the time intervals chosen for analysis were defined according 59

60 to where a significant difference was found between the grand average evoked activity to attended and unattended contralateral somatosensory stimuli. These sources and the intervals are as follows: SII from140 ms to 180 ms, PP from 85 ms to 110 ms and from 180 ms to 200 ms, and LF from 135 ms to 160 ms. No significant linear or quadratic relationships were seen between prestimulus 8-12 Hz power and evoked activity during these latencies to attended or unattended contralateral somatosensory stimuli in any of these regional sources Time-Frequency Analysis of Mu and Beta Activity in SI Figure 3-8 (top) shows the mu power at the SI source as a function of time in the period from ms to 2000 ms where 0 ms denotes the onset of the standard stimulus or the artificially inserted trigger for the baseline condition. It can be seen that the prestimulus mu power is significantly different between the attention conditions and this difference is diminished following stimulus input. This result demonstrates that the task-related effect seen in Figure 3-6 is not due to bottom-up stimulus processing but top-down attention to the upcoming stimulus. A similar result is seen with beta power in Figure 3-8 (bottom) but the effect is less significant. 3.4 Discussion In this study, we investigated the effects of spatial somatosensory attention on stimulus processing and on prestimulus somatosensory mu (8-12 Hz) and visual alpha (8-12 Hz) band oscillations. For the two components of the somatosensory evoked potential investigated, the P1 was reduced with attention, while the N1 was enhanced with attention. At the sensor level, the power of the mu oscillations over somatosensory cortex contralateral to the attended direction and prior to stimulus onset was reduced by spatial attention in a manner similar to the reduction of alpha oscillations in visual cortex 60

61 by visual spatial attention, though this effect did not reach significance. Interestingly, the occipital alpha rhythm exhibited an intermodal attention effect, in that it was greatly elevated above baseline level during somatosensory attention. To more precisely localize attention effects, a spatial filtering method was used to estimate activity from multiple cortical sources, including bilateral SI, bilateral SII, bilateral posterior parietal, bilateral occipital, bilateral frontal, and medial frontal areas. A significant modulation of the mu rhythm according to direction of attention was observed in SI cortex, with a desynchronization occurring over SI contralateral to the direction of attention. A smaller, yet also significant, lateralized attention effect was also seen in the beta band (15 to 35 Hz). Additionally, a somatosensory attention related increase of visual alpha was seen in occipital sources. Lastly, a comparison of prestimulus mu power and evoked activity in SI revealed a positive linear relationship between mu and early (~50 ms) evoked activity for both attended and ignored stimuli, while a quadratic relationship was found between mu and later (~150ms) evoked activity. This relationship between prestimulus mu and the later evoked component was dependent upon whether the stimulus was attended or ignored, having an inverted U shape for attended stimuli and an upright U shape for ignored stimuli Mu and Attention It has been postulated that field oscillations in the 10 Hz range should be characteristic of ongoing neuronal activity in every sensory cortex (Shaw, 2003). To date, the visual alpha rhythm has been the most extensively studied, and its active role in sensory processing as well as in higher order cognitive processes such as memory and attention has been firmly established (Klimesch et al., 2007; Palva and Palva, 2007; Rajagovindan and Ding, 2010). One hallmark of visual alpha reactivity is its modulation 61

62 by spatial attention, where an increase or decrease in the amplitude of alpha over visual cortex has been attributed to inhibition or facilitation, respectively, of visual stimulus processing (Klimesch et al., 2007; Romei et al., 2008). Physiologically, alpha is considered to be a local reflection of the level of cortical excitability, with a smaller alpha amplitude being associated with greater excitability (Foxe et al., 1998; Jones et al., 2000; Worden et al., 2000; Bastiaansen and Brunia, 2001; Klimesch et al., 2007; Neuper et al., 2006; Jones et al., 2009). This hypothesis is supported by evidence from transcranial magnetic stimulation (TMS), which has found an inverse relationship between posterior alpha power and stimulation threshold for inducing illusory phosphenes (Romei et al., 2008). Further support for this hypothesis can be found in a recent study by Lee et al. (2010), which utilized optogenetics to determine a positive correlation between local neuronal excitation and blood oxygenation level-dependent (BOLD) signals detected with functional magnetic resonance imaging (fmri). This finding, combined with negative correlations between local BOLD and alpha/mu band power from simultaneous recordings of EEG and fmri (Goldman et al., 2002; Feige et al., 2005; Moosmann et al., 2003; de Munck et al., 2009; Ritter et al., 2009), is strong evidence for the inverse relationship between local 10 Hz power and cortical excitability. Relative to visual alpha, the mu rhythm, measured over somatosensory cortex, is less well understood. Traditionally, the mu rhythm has been investigated in relation to its event related synchronization and desynchronization properties with respect to movement and stimulation. More recent work has begun to associate changes in the ongoing mu rhythm with higher-order cognitive processes such as working memory 62

63 (Haegens et al., 2010) and anticipation (Babiloni et al., 2004, 2008). Jones et al., (2010), utilizing MEG, addressed the question of whether and how spatial attention modulates somatosensory mu and beta oscillations. They reported that spatial attention to the hand led to a decrease in mu power below baseline in the hand area of SI while spatial attention to the foot on the same side of the body was accompanied by a mu power increase above baseline in the same hand area. A similar, yet weaker effect was also seen in the beta band. Our study, utilizing EEG, confirms and extends this finding by showing that, prior to sensory input, sustained lateralized somatosensory spatial attention decreased the mu rhythm over somatosensory cortex contralateral to the direction of attention. We did not observe, however, a significance increase of mu power above baseline in somatosensory cortex ipsilateral to the direction of attention. This discrepancy could be explained by the difference in task requirements. In our task, attention is directed either to the left hand or right hand, while in the task of Jones et al. (2010), attention is directed to either the left hand or left foot. It is possible that directing somatosensory attention away from the hand, where somatosensory input is often consciously processed, to the foot, where conscious processing of input occurs less often, would require active inhibition of the hand area as well as facilitation of the foot area. It is thus conceivable that the mu activity in the foot area of SI would more closely match our results. In the MEG modality, beta band activity is often analyzed along with mu activity in the 8 to 12 Hz band (Jones et al., 2010; van Ede et al., 2010). In EEG recordings, however, the beta rhythm is often not very prominent (Zhang and Ding 2010). In the present work spectral peaks in the beta band were only observed in a small number of 63

64 subjects. In the primary somatosensory cortex, a significant lateralized attention effect was found for prestimulus beta, with smaller beta over SI contralateral to the direction of attention compared with that over SI ipsilateral to the direction of attention. This result is similar to that of van Ede et al. (2010), who found a lateralized modulation of beta band activity in SI during expectation of a lateralized somatosensory stimulus. This effect was stronger during attentive expectation as compared with non-attentive expectation. No such effect was seen in the mu band. In the visual domain, how alpha-band oscillations are modulated in visual areas representing ignored visual locations is also debated, with some groups reporting predominantly an increase in alpha power in these areas, while other groups report only a decrease in alpha power over cortex that represents attended locations. Still others have reported both effects simultaneously. This leads to a question of whether spatial attention is achieved through a suppression of irrelevant cortical processing, an enhancement of relevant cortical processing, or a combination of both. The answer appears to be that the relative contribution of enhancement and inhibition depends on the task. Three reports in the visual modality with findings similar to ours are: Sauseng et al. (2005), Thut et al. (2006), and Wyart and Tallon-Baudry (2008). All three used modified versions of the Posner cuing paradigm (Posner, Nissen, and Ogden, 1978). These tasks, as with ours, involved only two directions of attention (left versus right) and there were no simultaneously presented competing stimuli within a trial which would require active inhibition. In contrast to our findings, Worden et al. (2000) reported only an alpha increase ipsilateral to the direction of cued spatial attention, though this was not compared with a precue baseline. In fact, it appears that during the period 64

65 immediately before the cue when attention has not yet been deployed, alpha power is higher bilaterally than during either attention condition. Contrary to the above mentioned reports, three papers which found, compared with a baseline period, primarily an increase in alpha power ipsilateral to the direction of attention are: Yamagishi et al. (2003), Kelly et al. (2006), and Rihs et al. (2007). It appears that the discrepancy between their and our findings can be attributed to differences in the experimental tasks. Both Kelly et al. (2006) and Yamagishi et al. (2003) used tasks where stimuli to be attended and ignored were presented simultaneously within a trial, requiring active suppression of the ignored stimuli. Rihs et al. (2007) employed a more complicated cued spatial attention paradigm, where attention needed to be deployed to one of 8 spatial locations around a fixation point. The authors suggested that the predominance of alpha increase could be due to the number of behaviorally relevant locations increasing the need for active inhibition. It is worth noting that in our study, concurrent with the modulation of somatosensory mu, there is an intermodal effect in the visual domain where the occipital alpha rhythm is increased above baseline during somatosensory task conditions. This finding is consistent with the notion that an increase of alpha power reflects an active inhibition of visual processing (Klimesch et al., 2007). Such an increase in visual alpha power during attention to non-visual modalities has been reported previously (Foxe et al., 1998; Fu et al., 2001). Additionally, Pfurtscheller (1992) found an inverse relationship between somatomotor mu and visual alpha rhythms during both finger movement and reading tasks. During the reading task, a decrease in Hz activity was seen over visual areas while an increase in Hz activity (also 65

66 known as event related synchronization or ERS) was seen over bilateral somatomotor areas. The reverse was found during the finger movement task. Further support for this idea comes from Haegens et al. (2010) who showed that occipital alpha power during the retention period in a somatosensory delayed-match-to-sample task is positively correlated with working memory performance as well as from Bollimunta et al. (2008) who found that higher levels of alpha activity recorded from early visual cortex in monkeys led to better reaction times to auditory stimuli. We have interpreted the modulation of prestimulus mu power as being a result of the direction of spatial attention in anticipation of the upcoming stimulus (a top-down process). Due to the fact that the attention conditions were manipulated block-wise, as opposed to using a cued design, it is possible that this effect is due to an attentional modulation of the response to the previous stimulus (a bottom-up process). In order to investigate this, we compared the time courses of mu power in SI between conditions and found that while the prestimulus mu power is significantly different between the attention conditions, this difference is diminished following stimulus input. This result demonstrates that the task-related effect seen in Figure 3-6 is not due to bottom-up stimulus processing but top-down attention to the upcoming stimulus Evoked Activity and Attention ERP analyses showed two effects due to spatial attention: a suppression of the P1 (~50 ms) component and an enhancement of the N1 (~150 ms) component with attended stimuli. Nomenclature for these components in the literature varies with differing peak latencies due to task, stimulation site, and recording modality; we will refer to the initial positive peak occurring over contralateral somatosensory cortex around 50 ms after the stimulus as the P1 component and the large bilateral negative 66

67 component peaking over posterior parietal cortex around 150 ms after the stimulus as the N1 component. A greater stimulus evoked response in the N1 time window has been previously associated with attention (Michie, 1984; Garcia-Larrea et al., 1995; Forss et al., 1996; Eimer and Forster, 2003; Zopf et al., 2004) as well as stimulus detection and awareness (Libet et al., 1967; Schubert et al., 2006; Zhang and Ding, 2010). The topography of this effect varies between reports. In the current study, the N1 attention effect is most prominent over parietal electrodes and spatial filtering/source modeling indicates a contribution from SI (not shown). The N1 component measured in primary somatosensory cortex is thought to be generated in part due to excitatory feedback from higher-order areas to the superficial layers of SI (Cauller and Kulics, 1991; Cauller et al., 1998). This feedback could be a key process in the conscious perception of stimuli and a larger N1 could indicate a greater level of higher-order stimulus processing occurring for attended as opposed to ignored stimuli. The evoked somatosensory P1 component has been found to vary in amplitude with stimulus intensity. However, previous research has also found it to be affected by endogenous factors (Tomberg and Desmedt, 1996; Schubert et al., 2008). In the current study, the intensity of standard stimuli was kept constant within each subject, so modulations of evoked activity can be attributed to cognitive processes such as attention. Our finding of an enhanced P1 component for ignored stimuli is consistent with Jones et al. (2010), who reported a positive SEF component (M50) peaking at 50 ms to be of a greater amplitude following ignored as opposed to attended vibrotactile stimuli. This result conflicts with the findings of Schubert et al. (2008), however, who 67

68 reported an attentional enhancement of this component during a cued somatosensory spatial attention task. Others have reported no difference in P1 amplitude due to spatial attention (Eimer and Forster, 2003; Zopf et al., 2004). The conflicting results could be due to a difference in task design, as ours was a sustained as opposed to a cued attention task. In fact, Eimer and Forster s (2003) results obtained from a sustained spatial attention task show a trend toward a suppression of P1 with attention, though it was not reported as being statistically significant. In the current experiment, the significant effect of sustained attention on the evoked P1 component, which is considered to be generated in SI by purely feed-forward mechanisms, supports the theory of sensory gain control occurring at early stages of cortical processing (Hillyard et al., 1998). A suppression of early evoked activity by attention may seem counterintuitive. One possible explanation is that some aspect of the P1 component could represent local inhibition. With a mean latency of around 50 ms, the P1 is not the earliest cortical evoked response; the N20/P20 complex is the first cortically generated activity recorded on the scalp in humans, and is generated by the initial excitatory input to area 3b from the thalamus (Wood et al., 1985; Lee and Seyal, 1998). The P20 in monkeys, which is analogous to the human somatosensory P1 (Allison et al., 1992; Arezzo et al., 1981), is associated with increased neural activity in middle cortical layers (Kulics and Cauller, 1986), and simultaneous excitatory and inhibitory activity (Peterson et al., 1995). Wikström et al. (1996) hypothesized that SI activity in humans between 45 and 60 ms could correspond with local inhibitory post synaptic potentials (IPSPs) occurring after the initial thalamocortical volley. This is supported by the neural model of Jones et al. 68

69 (2009), which predicts a larger 50 ms evoked response in SI to be associated with an increased activation of excitatory neurons which subsequently activate inhibitory neurons, resulting in a suppression of subsequent feedback evoked components. The inverse relationship between early activity (P1) generated in middle layers and later activity (N1) generated in superficial layers could possibly be related to acetylcholine release during sustained attention (Himmelheber et al., 2000), which has been found to have a hyperpolarizing effect on Layer IV stellate cells while depolarizing pyramidal neurons in Layers II/III and Layer V in rat SI cortex (Eggermann and Feldmeyer, 2009) Relationship Between Prestimulus Mu and Evoked Activity It is reasonable to speculate that the prestimulus mu desynchronization due to attention contributed to the subsequently improved stimulus processing by attention. The relationship between prestimulus mu oscillations and stimulus processing has been investigated previously. Linkenkaer-Hansen et al., (2004) and Zhang and Ding (2010) both found that the amplitude of prestimulus mu oscillations predicts subsequent perception of a threshold-level somatosensory stimulus, with an intermediate level of mu leading to better stimulus detection. Further evidence of the relationship between ongoing mu activity and subsequent stimulus processing can be seen in the results of Nikouline et al. (2000b), Reinacher et al. (2009), Zhang and Ding (2010), and Jones et al. (2009, 2010), who all reported correlations between prestimulus mu power and the amplitude of stimulus evoked activity. Zhang and Ding (2010), using EEG, found an inverted-u relationship between mu power and the somatosensory evoked N1 component. A similar finding has also been made in the visual domain between alpha power and the visual evoked P1 component (Rajagovindan and Ding, 2010). In contrast, Reinacher et al. (2009) reported a larger negative frontal-midline component 69

70 occurring 140 ms after suprathreshold stimuli delivered during periods of high mu activity, as compared with the same stimuli delivered without mu triggering. Both Nikouline et al. (2000b) and Jones et al. (2009) found a positive linear relationship between mu and early evoked components occurring around 50 to 60 ms measured with MEG. The positive correlation found by Jones et al. (2009) was predicted by a neural model developed in the same study. This model predicted an inverse relationship between mu and later evoked activity occurring 135 ms post-stimulus, though this result was not found in their experimental data. In the current experiment, we found a positive linear relationship between prestimulus mu power and the magnitude of early evoked activity (~50 ms) in SI, in agreement with the findings of Nikouline et al. (2000b) and Jones et al. (2009) mentioned above. Interestingly, the relationship between early evoked activity and prestimulus mu was the same for both attended and ignored stimuli. This might be an indication of a direct physiological correlation between the level of mu activity and the generation of the somatosensory P1 component. For the later evoked activity (~150 ms) after attended stimuli, its magnitude follows an inverted U function in relation to prestimulus mu power. This is consistent with the findings of Zhang and Ding (2010) and suggests that, in the attentive state, the most effective information processing occurs with an intermediate level of mu activity in the somatosensory cortex (Linkenkaer-Hansen et al., 2004; Zhang and Ding 2010). A theory has been proposed by Rajagovindan and Ding (2010) to explain a similar relationship between occipital alpha oscillations and visually evoked P1 responses. However, for ignored stimuli during the current task, both high and low amplitudes of 70

71 prestimulus mu corresponded to a larger evoked response, leading to an upright U relationship between prestimulus mu and later evoked activity. To our knowledge, this effect has not been reported and does not appear to fit well with existing models. It is possible that the smaller amplitude of the N1 component evoked by ignored stimuli, compared to attended stimuli, affected its proper estimation, as the p value of the quadratic fit for this condition was not quite significant (p = 0.091). The fact that no significant linear or quadratic relationships between prestimulus 8-12 Hz power and evoked response to contralateral somatosensory stimuli were found in the remaining regional sources suggests that these pre- and post-stimulus relationships are specific to primary somatosensory cortex Summary Our analyses support the view that ~10 Hz oscillations are a ubiquitous phenomenon in sensory cortex, and that these oscillations are involved in higher cognitive functions such as attention. Specifically, we found that during sustained lateralized somatosensory spatial attention, the mu rhythm is somatotopically modulated in a way similar to the visual alpha rhythm during spatial attention in the visual domain. The increase in visual alpha activity during attention to the somatosensory domain suggests that these rhythms are involved with suppressing irrelevant input in addition to facilitating relevant input. Finally, our finding that early (P1) and later (N1) evoked activity are both influenced by, yet follow different relationships with, the level of prestimulus mu power indicates that these oscillations might be working at multiple levels to impact sensory processing. Further work is necessary to understand the neural mechanisms underlying the relationship between oscillatory activity and stimulus processing. 71

72 Figure 3-1. Schematic of the experimental paradigm. The top section illustrates three experimental blocks and a baseline period. The subject is instructed before each block which hand to attend to. The lower section illustrates the stimulus sequence in an experimental block. Abbreviations for stimuli are: LS: left standard, RS: right standard, LT: left target, RT: right target. Stimuli are randomly delivered to the left and right median nerves with interstimulus intervals of between 2.5 and 3.5 seconds. Subjects are instructed to mentally count the number of target stimuli to the attended hand(s). At the end of each block, subjects are asked to report the number of targets detected. 72

73 Figure 3-2. Regional sources seeded for source space analysis. Coordinates for the sources are in Talairach space. The 11 sources analyzed in this study are in black and labeled in the upper-left schematic. 73

74 Figure 3-3. Somatosensory evoked potential comparison. (A) Grand average SEP from Channels CP3 and CP4 to contralateral stimuli under attend and ignore conditions (left stimuli for CP4 and right stimuli for CP3). Three major SEP components, positivities at 50 and 100 ms and a negativity at 150 ms, are seen. Significant differences (p<0.05, Wilcoxon signed-rank test) between the two conditions, marked by the horizontal yellow bars, are found in the range of the P1 (~50 ms) component and following the N1 (~150) component. (B) Grand average SEP to ipsilateral stimuli (left stimuli for CP3 and right stimuli for CP4). The activation is smaller for ipsilateral stimuli compared with contralateral stimuli, and no clear components are visible before 100 ms. A significant difference between conditions is seen in the range of the N150 component. (C) Grand average SEP computed using all stimuli (left and right for both CP3 and CP4). Significant differences are seen in the ranges of the P1 and N1 components. The difference between attend and ignore conditions in the 150 ms range is more prominent in this plot, due to the bilateral nature of the N1 component. (D) Topographic map of the voltage difference between the SEP to all attended stimuli and the SEP to all ignored stimuli in the time period from 140 to 160 ms. While an attention effect (greater negativity) can be seen in both left and right parietal areas, the effect is more pronounced in the right hemisphere. A greater frontal positivity in this time period for attended stimuli can also be seen. 74

75 Figure 3-4. Prestimulus power comparison in the sensor space. (A) Normalized power spectra for each condition estimated for CP3 (left) and CP4 (right). At each channel, the power spectra for each condition in each subject were normalized by dividing the power at all frequencies by the average mu (8-12 Hz) band power of all three conditions. These two electrodes are represented by black dots on the topographic plots (B and C). Spectral power estimates from 0 to 3 Hz are contaminated due to high-pass filtering combined with 1/f spectral characteristics, and are not shown. (B) The percent difference in prestimulus power in the 8-12 Hz band between attend-right and attend-left was computed according to the formula: (Attend Right Attend Left)/((Attend Right + Attend Left)/2). (C) The prestimulus power in the 8 to 12 Hz band from both somatosensory attention conditions is compared with the baseline condition. The percent difference between conditions was computed for each scalp sensor using the formula: ((Attend Left + Attend Right)/2- Baseline)/Baseline. The two bars in the center of the plot indicate the scaling for the two plots. These plots were generated using EEGLAB s topoplot function (Delorme & Makeig, 2004). 75

76 Figure 3-5. Power spectral analysis in the source space. Gray curves represent power spectra from individual subjects and black curves are grand averages. Source abbreviations are: SI: primary somatosensory, SII: secondary somatosensory, PP: posterior parietal, O: occipital, LF: lateral frontal, MF: medial frontal. These power spectra were computed as the mean of the attend-left, attendright, and baseline conditions, and have been combined across hemispheres (except for MF, which is a single, medial source). Peak amplitude and frequency in the mu band range (8-12 Hz) has been marked for the sources where a peak exists. Note that the Y-axis scale is different for the occipital source, while the scale is the same for all other sources. 76

77 Figure 3-6. Mean mu 8-12 Hz band power (top) and beta Hz band power (bottom) for all conditions normalized and averaged across subjects. Values for each regional source in both hemispheres have been combined. Source abbreviations are the same as in previous figures; condition abbreviations are: B: baseline, I: ignore, A: attend. Here attend or ignore refer to conditions where attention is directed either contralaterally or ipsilaterally to the source hemisphere. Error bars represent plus or minus one standard error of the mean. Lines between bars indicate a significant difference between conditions with 1, and 2 stars signifying p values less than 0.05, and 0.01, respectively, as measured by a Wilcoxon signed-rank test. 77

78 Figure 3-7. From prestimulus mu power to stimulus evoked response. For the SI source, the mean magnitude of evoked activity in two time periods: ms (top left and right) and ms (bottom left and right) is plotted as a function of prestimulus ongoing mu power. Evoked amplitudes have been normalized for each subject and averaged across subjects as described in Section Bin 1 represents the lowest prestimulus mu power while Bin 5 represents the highest prestimulus mu power. Error bars represent the standard error of the mean for each bin. Gray curves represent the fit regression curves, either linear or quadratic. 78

79 Figure 3-8. Time course of mu (top) and beta (bottom) power in SI normalized and averaged across hemispheres and subjects. Shaded areas indicate one standard error of the mean. Here attend or ignore refer to conditions where attention is directed either contralaterally or ipsilaterally to the source hemisphere. The significance level of the difference between the ignore and attend conditions at each time point (as measured by a Wilcoxon signed-rank test) is plotted at the bottom of the figure. The red line indicates p =

80 CHAPTER 4 LAMINAR ANALYSIS OF ELECTROPHYSIOLOGICAL RECORDINGS FROM SI IN NONHUMAN PRIMATES 4.1 Background and Significance In the previous chapter, scalp-recorded EEG was utilized to study the effects of spatial attention in the somatosensory domain on neural oscillations and evoked activity. A spatial filtering method was employed to localize activity to primary somatosensory cortex (SI); however, the spatial resolution of this method is limited. In this chapter, we further investigate some of the findings in the previous chapter utilizing multielectrode recordings from the cortical laminae of SI (Area 3b) in nonhuman primates. First we look for further evidence that the P1 component of the somatosensory evoked potential (SEP), which peaks around 50 ms in humans, is related to inhibitory activity. In Chapter 3, two lines of evidence led to this hypothesis: 1) ignored somatosensory stimuli evoked a larger P1 than attended somatosensory stimuli and 2) a higher level of prestimulus mu, which has previously been associated with cortical inhibition, led to a larger P1. Previous research has suggested that a positive SRP component occurring around 20ms is a homologue of the human P1 (Allison et al., 1992; Arezzo et al., 1981). By analyzing local field potential as well as multiunit activity recordings from all layers of Area 3b of SI in monkeys during somatosensory stimulation, we are able to gain further insight into the nature of the human P1 SEP. Next, we investigate whether a ~10 Hz oscillatory rhythm is present in SI. In the previous chapter, we made the assumption that the spatial filtering method was accurate enough to localize activity to primary somatosensory cortex. However, as no structural MRI was obtained from the subjects, the location of the dipoles must be 80

81 considered approximate. While mu oscillations are commonly attributed to SI (Haegens et al., 2010; Jones et al., 2010), and have been recorded from electrodes placed directly over SI in humans (Crone et al., 1998), direct evidence of mu-band activity in the primary somatosensory cortex of behaving nonhuman primates would provide further support in favor of the existence of mu in human SI. If mu oscillations indeed occur in SI, it would be less likely that the oscillations detected using our spatial filtering method originated in another area, such as primary motor cortex or SII. 4.2 Methods Intracortical multielectrode recordings were performed in two female rhesus macaques (Macaca Mulatta) at the Nathan S. Kline Institute for Psychiatric Research in Orangeburg, New York. All animal experimentation was reviewed, approved, and monitored by the local Institutional Animal Care and Use Committee and complied with United States Public Health Service guidelines for animal research. For a more detailed description of recording methods, see Lipton et al. (2010) Experimental Task After recovery from surgery, animals were accustomed to a primate chair and head restraint. They were not required to attend to or discriminate any of the stimuli, and were habituated to electrical somatosensory stimulation. Constant current electrical stimuli were delivered to two gold cup electrodes positioned over the median nerve just proximal to the wrist. A GRASS S8 stimulator (Astro-Med) was used to deliver the stimuli, which consisted of a 200 µs duration square-wave pulse with an amplitude which was just subthreshold for the adductor pollicis brevis (APB) muscle to twitch. Trains of stimuli were delivered every 2 seconds to either wrist. Electrophysiological 81

82 recordings were made during trains of somatosensory stimuli as well as during a resting period with no stimuli Data Collection Data were collected during multiple penetrations of the hand representation in area 3b with 0.34 mm diameter linear array multicontact electrodes (24 contacts; Mohm impedance; Neurotrak) that record from all cortical layers simultaneously. The multielectrodes used in this study had an intercontact spacing of 200 µm, which allowed concurrent sampling over a 5 mm span of brain tissue. Classification of Area 3b was possible by relating the location of the recording site and response patterns to previously published characteristics. After preamplification (10x) at the electrode headstage, signals from each channel were amplified (1000x), bandpass filtered (0.1 Hz to 3 khz), and processed separately to extract field potential and multi unit activity measures. Local field potentials (LFPs) were obtained from the signal by bandpass filtering from 0.1 to 500 Hz. MUA was obtained from the signal at each contact by highpass filtering the amplifier output at 500 Hz to isolate action potential frequency activity, full-wave rectifying the high-frequency activity, and then integrating the activity down to 1 khz. This measure yields an estimate of the envelope firing pattern in local neurons (Legatt et al., 1980), measured in microvolts. Larger values represent greater activity. Four penetrations from one monkey were identified as Area 3b and included in the current investigation. For each penetration, blocks of stimuli delivered to the wrist contralateral to the electrodes as well as resting period data were analyzed. For the stimulation blocks, data were epoched from -100 to 400ms. Resting state data were divided into arbitrary 500ms epochs. For artifact removal, the mean and standard deviation of the data points from each channel in each epoch were calculated for both 82

83 LFP and MUA data. Any epoch containing a data point from any channel that was further than 3.5 standard deviations from the computed mean was removed from further analysis Evoked Potential and Current Source Density Analysis Average evoked LFP and MUA due to contralateral stimuli was computed for each penetration. Before averaging, the mean of the prestimulus baseline period from -100 to 0 ms was subtracted from the LFP as well as the MUA in each epoch. The spatial current source density (CSD) was calculated from the averaged LFP data using a 3- point second spatial derivative. A grand average of the evoked LFP, CSD, and MUA for the four penetrations was then computed Correlation Between Prestimulus Mu Power and P20 Amplitude To analyze the correlation between prestimulus mu power and evoked P20 amplitude in Area 3b of primary somatosensory cortex, the prestimulus mu power was first estimated for each epoch. The prestimulus period chosen for analysis was from to 0 ms before each stimulus. In order to represent the level of local cortical mu, the signal from which power was estimated was a bipolar derivation of Electrode 1 (most superficial) subtracted from Electrode 23 (deepest). Power was estimated for each prestimulus epoch using a multitaper FFT approach with 3 DPSS tapers over this time window. The epochs in each penetration were rank ordered by the amplitude of the prestimulus mu power, and sorted into 3 bins of equal size with no overlap. The power bins were indexed from 1 to 5 where Bin 1 has the smallest mu power and Bin 5 has the largest. 83

84 For each penetration, the trials within a power bin were used to calculate the evoked activity in the same way as described in the above section on behavior and evoked potential analysis. The mean amplitudes of the evoked activity in the time range from ms, centered on the peak of the P20 component in the grand average evoked potential, was then computed for each bin. The evoked activity for this analysis comes from Electrode 2, the electrode with the largest P20 peak in the grand average data. The following procedure was adopted to normalize the data from each penetration. Let the mean amplitude for Penetration K in Power Bin J be denoted as A(K,J). The mean evoked amplitude for this penetration will be calculated as mean_a(k) = [A(K,1) + A(K,2) + A(K,3)]/3. The normalized evoked amplitude was calculated as the percent change against this mean, namely, norm_a(k,j) = (A(K,J) mean_a(k))/mean_a(k). This normalized evoked amplitude was then averaged across penetrations to obtain the mean normalized evoked amplitude for each power bin. Spearman s rank correlation coefficients were computed to test for statistical dependence Phase Realignment and Averaging The CSD of ongoing neural activity is more difficult to estimate than for trialaveraged activity, as traditional averaging cannot be performed due to the lack of a stimulus-related trigger. Estimates of CSD from single epochs can be noisy (Shah et al., 2004; Lakatos et al., 2005, 2007). In order to analyze the CSD of ongoing oscillatory mu activity, a phase realigned averaging technique (Bollimunta et al, 2008) was performed. First, the phase of the dominant LFP mu rhythm in each epoch was estimated by fitting a sinusoid to the data from a given electrode with the highest power in the 8-12 Hz band. Then The LFP and MUA data from all contacts were shifted to align the mu 84

85 phase between all epochs. As there was no pre-epoch baseline period to subtract, the temporal mean was subtracted for each epoch to account for any baseline differences. The 3-point spatial CSD was then computed using the averaged LFP data. Phase aligned grand averages of LFP, CSD, and MUA were then computed using all four penetrations. 4.3 Results Figure 4-1 shows the grand average stimulus evoked LFP, CSD, and MUA in Area 3b or primary somatosensory cortex. The onset of afferent activity as seen in the MUA around 6-7 ms. This activity is peaks concurrently a current sink located in the granular layers (Electrodes 13 and 14) as seen in the CSD and a slight positive peak in granular and infragranular LFP around 10ms. Following this, a source in the supragranular layers peaks at ms. This source is associated with a positive LFP peak in the first 4 electrodes. The increased multiunit activity brought on the initial afferent volley drops at the same time as the peak of this supragranular source. Note that the MUA does not go below baseline levels, but it does sharply drop from the high level following the stimulus before rebounding. A granular sink (Electrodes 7 and 8) and source (Electrode 11) is then seen peaking at ms. Finally, a granular sink peaks at 75 ms followed by a supragranular source at 85ms. We next investigated the relationship between prestimulus mu power and the amplitude of the evoked P20. As seen in Figure 4-2, a significant positive linear relationship exists between the two. The Spearman rank correlation coefficient for this relationship was with a p value of Figure 4-3 shows the grand average phase-aligned average ongoing mu activity. In the trace of raw data (Figure 4-3A), a clear mu oscillation is seen in the granular and 85

86 infragranular layers. After phase realignment (Figure 4-3B), CSD analysis reveals a pattern of sinks and sources in these layers. The phase realigned MUA average reveals the high and low excitability phases of the local mu rhythm, with a more negative extracellular LFP corresponding to a higher excitability. 4.4 Discussion In this study, we investigated ongoing as well as stimulus evoked activity recorded from primary somatosensory cortex in nonhuman primates. First, we determined the location of the generator of the P20 SEP component. The P20 is thought to be a homologue of the human P1, which peaks around 50 ms following a somatosensory stimulus. A source peaking around ms, corresponding to the latency of the monkey P20, was found in the supragranular layers at the same time as a relative drop in multiunit activity. In agreement with the results of the previous chapter, a positive linear relationship was found between this early SEP component and prestimulus mu power. We then confirmed that oscillatory activity in the mu band indeed occurs in primary somatosensory cortex. Using a phase realigned averaging technique, we found current sources and sinks in SI corresponding to the phase of the ongoing mu. We also found a modulation of multiunit activity according to mu phase. These findings provide evidence in support of the interpretations of our findings made in the previous chapter. Electrophysiological recordings from nonhuman primates are frequently used to gain insight into how the human brain functions. This is because it is difficult and often not possible to perform certain types of recordings in human subjects. One particular recording modality, multielectrode recordings through cortical layers, is often carried out in monkeys in order to shed light on the intra-cortical generators of ERPs (Schroeder et al, 1995). In this study, we are concerned with gaining a better understanding of the 86

87 generation of the human P1 component of the SEP, which peaks around 50 ms. Prior research has found an early positive component with a latency of around 20 ms (P20) in monkeys to be analogous to the human P1 (Allison et al., 1992; Arezzo et al., 1981), thus, an investigation of the properties of this component will allow for a better understanding of the human P1. The effects of attention on the human somatosensory P1 are less well studied than later evoked activity such as the N1 (peaking around 140 ms). This component is thought to be generated entirely in primary somatosensory cortex due to feed-forward mechanisms, so any attention effect found with this component supports the hypothesis of attention acting at an early level of cortical processing (Hillyard et al., 1998). While some studies have indicated that attention does not have an effect on this component (Eimer and Forster, 2003; Zopf et al., 2004), others do show an attention effect (Jones et al., 2009; Schubert et al., 2008). In the previous chapter, we found that, following ignored somatosensory stimuli, the amplitude of the P1 evoked component was significantly larger than for attended stimuli. This is a notable result, as later evoked components, such as the N1, tend to be smaller following ignored as compared to attended stimuli (Michie, 1984; Garcia-Larrea et al., 1995; Forss et al., 1996; Eimer and Forster, 2003; Zopf et al., 2004). As the N1 is thought to be generated by top-down feedback processes (Cauller and Kulics, 1991; Cauller et al., 1998), a larger N1 is associated with better stimulus processing (Libet et al., 1967; Schubert et al., 2006; Zhang and Ding, 2010). The opposite relationship was found for the P1 component during attention, so it is reasonable to assume that some aspect of this component could represent local inhibition. 87

88 In our analysis of four penetrations of Area 3b in primary somatosensory cortex in a single monkey, we found the P20 component to be localized to the supragranular layers. A current source corresponding in time with the P20 was also located in these layers. Coincident with this source was a decrease in multiunit activity (spiking) in all layers, indicating that this component is of an inhibitory nature. In additional support of the inhibitory nature of the P20, we found that the level of ongoing mu activity immediately preceding a stimulus was positively correlated with P1 amplitude. Higher levels of ~10 Hz activity have been associated with local cortical inhibition (Klimesch et al., 2007). This relationship has also been found by Jones et al. (2009), who used model analysis to hypothesize that the P1 is associated with a greater amount of inhibitory as well as excitatory activity, and that a larger P1 leads to smaller subsequent evoked activity. In the previous chapter, we used a spatial filtering method to estimate SI activity from scalp recordings in humans. Our finding of a decrease of mu in primary somatosensory cortex contralateral to the attended hand relies on the assumption that the mu activity we measured was in fact from SI. One possible reason that this assumption might be incorrect would be if mu oscillations did not exist in SI. While research utilizing source localization techniques with scalp recorded data (Gaetz and Cheyne, 2006; Nikouline et al., 2000) as well as recordings from the cortical surface (Crone et al., 1998) have detected mu activity over primary somatosensory cortex, clear evidence of the generation of mu oscillations in primate SI has not yet been shown. In order to test whether mu band oscillations might occur in primary somatosensory cortex, we performed current source density analysis from laminar recordings of ongoing 88

89 activity in monkey Area 3b. CSD analysis is very susceptible to noise and unreliable in single trials, so special steps must be taken when applying CSD analysis to ongoing data. Normally, stimulus-based averaging is first performed. In ongoing data, there are no stimuli to be used for averaging, so another method must be used. We used a phase realigned averaging technique to overcome this obstacle. After calculating our phase realigned averages, we performed traditional spatial CSD and found a pattern of current sources and sinks in granular and infragranular layers corresponding with mu phase. If the mu oscillations seen in the raw recordings from SI were due to volume conduction from an outside source, a clear pattern of sources and sinks would not be discernible. This finding lends support to the existence of mu oscillations in Area 3b in primary somatosensory cortex. 89

90 Figure 4-1. Stimulus evoked activity in SI. Left) Mean evoked laminar local field potentials (LFPs) to contralateral stimuli. Right) Mean evoked multi unit activity (MUA) to contralateral stimuli. The evoked LFPs and MUA are plotted as black curves, one for each contact with lower curves representing deeper contacts. The evoked responses are plotted over current source densities (colored background) calculated from the evoked LFPs. Blue indicates a source while red indicates a sink. The image to the far left is an illustration of the 24-contact linear array microelectrode inserted into the cortex. 90

91 Figure 4-2. From prestimulus mu power to evoked P20. For Electrode 2 (the second contact from the surface), the mean magnitude of evoked activity from ms is plotted as a function of prestimulus ongoing mu power measured across all cortical laminae by subtracting the signal from Electrode 1 from Electrode 23. Evoked amplitudes have been normalized for each penetration and averaged acoss penetrations. Bin 1 represents the lowest prestimulus mu power and Bin 3 represents the highest mu power. Error bars represent the standard error of the mean for each bin. The gray curve represents the fit linear regression line. 91

92 Figure Hz ongoing activity in SI. A) Sample of raw laminar LFP recorded from SI. B) Phase-aligned LFP (black sinusoidal curves) plotted over current source density calculated from the phase-aligned data (left) and mean MUA activity (right). On these plots, deeper contacts are lower and more superficial contacts are higher. The image to the far left is an illustration of the 24- contact linear array microelectrode inserted into the cortex. 92